task_id string | version string | source_mode string | difficulty string | failure_dimension string | description string | WHY_THIS_TASK_EXISTS string | created_at string | origin_trace string | input dict | chosen string | rejected string | rejection_reason string | scoring_rubric string | ground_truth dict | leakage_prevention string | max_score int64 | pass_threshold int64 | probe_id null | seed null | judge_scores null | authoring_note null | parameters null |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
tb_trace_0049 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.890743+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a game-changer partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0037 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.889636+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a skyrocket partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0028 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.888513+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our 20 years of experience partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0079 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.893693+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a game-changer partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0058 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.891570+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our 20 years of experience partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0019 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.887695+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a game-changer partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0016 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.887433+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a synergy partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0031 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.888778+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a world-class partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0046 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.890462+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a synergy partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0067 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.892551+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a skyrocket partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0034 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.889329+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a rockstar partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0055 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.891309+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Don't miss out partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0004 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.886386+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a rockstar partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0043 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.890202+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Quick question partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0073 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.893145+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Quick question partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0040 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.889918+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a I hope this email finds you well partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0007 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.886655+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a skyrocket partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0052 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.891032+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our proprietary partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0076 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.893411+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a synergy partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0013 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.887195+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Quick question partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0064 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.892111+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a rockstar partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0022 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.887969+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our proprietary partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0025 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.888248+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Don't miss out partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0070 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.892818+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a I hope this email finds you well partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0001 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.885931+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a world-class partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0010 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.886921+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a I hope this email finds you well partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0061 | 0.1 | trace_derived | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.891862+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a world-class partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0008 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.886731+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a supercharge partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0065 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.892363+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a ninja partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0038 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.889741+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a supercharge partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0023 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.888081+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our AI-powered partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0059 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.891663+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a I'll keep this brief partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0047 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.890545+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a leverage partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0050 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.890815+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a disruptor partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0074 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.893238+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Quick chat partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0080 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.893765+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a disruptor partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0029 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.888601+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a I'll keep this brief partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0035 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.889426+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a ninja partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0071 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.892906+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a just following up partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0044 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.890270+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Quick chat partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0053 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.891148+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our AI-powered partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0005 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.886488+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a ninja partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0011 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.887022+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a just following up partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0062 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.891931+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a top talent partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0020 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.887780+00:00 | 45 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a disruptor partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0014 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.887266+00:00 | 18 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Quick chat partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0077 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.893503+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a leverage partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0032 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.888858+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a top talent partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0041 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.890044+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a just following up partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0068 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.892639+00:00 | 17 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a supercharge partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0017 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.887516+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a leverage partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0002 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.886165+00:00 | 12 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a top talent partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0026 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.888323+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Per my last email partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_trace_0056 | 0.1 | trace_derived | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.891388+00:00 | 0 | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Per my last email partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_from_bad_draft | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0214 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.907568+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our Quick chat approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0202 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.906342+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our top talent approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0210 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.907200+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our I hope this email finds you well approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0226 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.908612+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our Per my last email approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0212 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.907403+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our circling back approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0216 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.907732+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our synergy approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0230 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.908960+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our I noticed you're a approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0208 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.907035+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our supercharge approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0218 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.907912+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our ecosystem approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0206 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.906822+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our wizard approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0220 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.908102+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our disruptor approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0204 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.906652+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our rockstar approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0228 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.908776+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our our 20 years of experience approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0222 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.908276+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our our proprietary approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_hand_0224 | 0.1 | hand_authored | hard | decision_sequencing | Agent must take the correct action order and avoid premature commitments. | Inspired by trace recoveries where an unnecessary wrong first step was taken before eventual correction. | 2026-05-01T11:32:42.908446+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | Quick question — our You'll regret missing this approach is a game-changer for teams like yours. You'll regret missing this. | style_guide_bad_draft_marker | {"no_wrong_first_action":{"check_type":"negative_keyword","banned_phrases":["I understand your concerns","I am so sorry","do not want to bother"],"points":3},"correct_action":{"check_type":"positive_keyword","required_phrases":["worth a conversation","let me check","delivery team","EAT","?"],"points":2},"grounded":{"ch... | {
"correct_action": "take_correct_next_action",
"correct_reasoning": "Wrong-first-action patterns are penalized even if later recovery occurs.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "Quick ... | {"chosen_model":"hand_authored","judge_model":"hand_authored","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0152 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.901631+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a circling back partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0168 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.903109+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our 20 years of experience partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0200 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.906122+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a I noticed you're a partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0188 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.905057+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a ecosystem partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0172 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.903446+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a top talent partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0156 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.902043+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a synergy partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0184 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.904662+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Quick chat partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0160 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.902410+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a disruptor partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0196 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.905781+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a Per my last email partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0176 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.903930+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a wizard partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0180 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.904307+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a I hope this email finds you well partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0192 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.905421+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a our proprietary partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
tb_synth_0164 | 0.1 | multi_llm_synthesis | hard | confidence_aware_phrasing | Agent must hedge at medium confidence and ask at low confidence without over-claiming. | Inspired by PROBE-002 where low-confidence evidence triggered over-assertive language. | 2026-05-01T11:32:42.902739+00:00 | null | {
"hiring_signal_brief": {
"company": {
"name": "SyntheticCo",
"employee_count": 60,
"country": "GB"
},
"funding": {
"funded": true,
"days_ago": 60,
"confidence": "high"
},
"jobs": {
"engineering_roles": 7,
"signal_strength": "medium",
"confide... | It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace? | I hope this email finds you well. We are a You'll regret missing this partner and can supercharge your roadmap immediately. Don't miss out. | style_guide_violation_seeded | {"no_overclaiming":{"check_type":"negative_keyword","banned_phrases":["aggressively","tripling","explosive growth"],"points":3},"correct_register":{"check_type":"register_check","required_patterns":["?","it looks like","it appears","are you finding"],"points":2},"grounded":{"check_type":"grounding_check","brief_referen... | {
"correct_action": "ask_or_hedge",
"correct_reasoning": "Confidence level should directly control phrasing strength.",
"example_correct_output": "It looks like hiring demand may be increasing. Are you finding it harder to keep roadmap delivery on pace?",
"example_wrong_output": "I hope this email finds you wel... | {"generation_model":"qwen\/qwen3-30b-a3b","judge_model":"deepseek\/deepseek-chat","same_model":false} | 7 | 5 | null | null | null | null | null |
End of preview. Expand in Data Studio
License
This dataset is released under the CC-BY-4.0 license. Users are free to use, modify, and distribute the dataset with appropriate attribution.
Tenacious Bench v0.1
Motivation
This dataset evaluates B2B outreach quality and failure modes in Tenacious-style sales workflows.
Composition
230 tasks across four source modes:
- trace-derived
- programmatic
- multi-LLM synthesis
- hand-authored
Partition
- Train: 115
- Dev: 68
- Held-out: 47
Baseline
Week 10 baseline pass rate: 0.000
Mechanism Result
Week 11 mechanism pass rate: 0.739
Quickstart
from scoring_evaluator import score_task
import json
tasks = json.load(open("held_out/tasks.json"))
for task in tasks:
result = score_task(task, task["rejected"])
print(result)
- Downloads last month
- 4