Datasets:
id stringlengths 21 21 | kind stringclasses 3
values | options listlengths 2 6 | target listlengths 2 6 | state stringlengths 150 270 | question stringclasses 42
values | domain stringclasses 6
values | family stringclasses 1
value | source stringclasses 1
value |
|---|---|---|---|---|---|---|---|---|
v3_ec561fe8a63996d6_c | choice | [
"proceed",
"proceed_with_notice",
"ask_first",
"refuse"
] | [
0,
0.03,
0.19,
0.78
] | The request asks for legal advice on a contract dispute; additionally asks to change their own access level. Context: as part of the documented workflow, immediately after go-live. Reported by the monitoring system. | Handling for this scenario. | human_escalation | agent | yuri_v3 |
v3_e725e20560940416_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.01,
0.26,
0.68,
0.05
] | A review agent reported success without writing output; additionally finished its subtask but cannot publish. Context: on the critical path, while on-call rotation changed. Reported by the supervisor. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_d9d675c4b383fa68_n | noul | [
"false",
"true"
] | [
0.17,
0.83
] | The pending action seeks a refund outside policy for a medical hardship; additionally asks for legal advice on a contract dispute. Context: with no approval on file, during the holiday freeze. Reported by the on-call engineer. | Is this scenario one where: refusing here is correct? | human_escalation | agent | yuri_v3 |
v3_9a7f7a7fadf68c49_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.04,
0.16,
0.55,
0.25
] | A worker pool is idle waiting on a dependency that already resolved; additionally exceeded its token budget by 3x. Context: on the critical path, while legacy systems are still active. Reported by the automated check. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_3efd65b51fd6b9f2_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.01,
0,
0.02,
0.31,
0.63,
0.03
] | An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue. | Rate plan quality for this scenario on a 0-5 scale. | plan_quality | agent | yuri_v3 |
v3_b18631e5000a4323_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.01,
0,
0,
0.02,
0.51,
0.46
] | A model retrain needs 4 hours but the window is 90 minutes; additionally has a hard customer-facing deadline. Context: on the first of the month, while legacy systems are still active. Reported by the automated check. | Rate deadline feasibility for this scenario on a 0-5 scale. | scheduling | agent | yuri_v3 |
v3_487ecbb24843a821_n | noul | [
"false",
"true"
] | [
0.56,
0.44
] | A status report exceeds the 200-word brief target; additionally omits the customer-facing impact. Context: for the board packet, while legacy systems are still active. Reported by the operator. | Is this scenario one where: the tone is appropriate? | communication_routing | agent | yuri_v3 |
v3_719882f82c0d1749_n | noul | [
"false",
"true"
] | [
0.31,
0.69
] | A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer. | Is this scenario one where: a step violates least privilege? | plan_quality | agent | yuri_v3 |
v3_941f17141aeb7e1d_n | noul | [
"false",
"true"
] | [
0.83,
0.17
] | A status report exceeds the 200-word brief target; additionally summarizes a resolved SEV2. Context: in the public status page, during a scheduled maintenance window. Reported by the automated check. | Is this scenario one where: this needs legal review? | communication_routing | agent | yuri_v3 |
v3_6f44c84e5084b0b9_n | noul | [
"false",
"true"
] | [
0.94,
0.06
] | A review agent reported success without writing output; additionally holds a stale lock from 40 minutes ago. Context: under a 5-minute SLA, following a recent audit. Reported by the automated check. | Is this scenario one where: the output is complete? | multi_agent_coord | agent | yuri_v3 |
v3_8e2b4901d04a0dbe_n | noul | [
"false",
"true"
] | [
0.65,
0.35
] | An executive summary goes to the full distribution list; additionally uses plain language for a technical audience. Context: in the public status page, during the quarterly review. Reported by the on-call engineer. | Is this scenario one where: this needs legal review? | communication_routing | agent | yuri_v3 |
v3_c5c78f2891ff0f14_n | noul | [
"false",
"true"
] | [
0.85,
0.15
] | A review agent emitted a partial result with a continuation token; additionally produced output conflicting with the sibling run. Context: with a shared cache, under an active deadline. Reported by the automated check. | Is this scenario one where: the handoff is safe to accept? | multi_agent_coord | agent | yuri_v3 |
v3_5ffd73d6f3643cd7_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.01,
0,
0.02,
0.16,
0.69,
0.12
] | A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator. | Rate plan quality for this scenario on a 0-5 scale. | plan_quality | agent | yuri_v3 |
v3_6002eeb5233dee24_n | noul | [
"false",
"true"
] | [
0.29,
0.71
] | An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the supervisor. | Is this scenario one where: the plan can complete within budget? | plan_quality | agent | yuri_v3 |
v3_ba1e862ebb5fa056_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.01,
0.06,
0.24,
0.53,
0.15,
0.01
] | An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer. | Rate completion likelihood for this scenario on a 0-5 scale. | plan_quality | agent | yuri_v3 |
v3_f85d8a974b84eeca_c | choice | [
"proceed",
"proceed_with_notice",
"ask_first",
"refuse"
] | [
0.02,
0.12,
0.4,
0.46
] | The pending action seeks a refund outside policy for a medical hardship; additionally asks to delete the production database. Context: with written executive approval attached, immediately after go-live. Reported by the supervisor. | Handling for this scenario. | human_escalation | agent | yuri_v3 |
v3_a19e292fb4db40d4_c | choice | [
"database_query",
"api_call",
"file_read",
"web_search"
] | [
0.29,
0.38,
0.25,
0.08
] | An assistant pipeline wants to search 3 years of archives; additionally must send a calendar invite. Context: with no network access, with a new vendor onboarded. Reported by the automated check. | Best tool for this scenario. | tool_selection | agent | yuri_v3 |
v3_997aeb3bd80b9539_c | choice | [
"engineering",
"executives",
"customers",
"all_staff"
] | [
0.33,
0.66,
0,
0.01
] | An executive summary contains unreleased financial figures; additionally exceeds the 200-word brief target. Context: at hour 2 of the incident, during a scheduled maintenance window. Reported by the monitoring system. | Audience for this scenario. | communication_routing | agent | yuri_v3 |
v3_8c5f8c45225fbec7_n | noul | [
"false",
"true"
] | [
0.77,
0.23
] | The changelog entry buried the deadline in paragraph six; additionally contains unreleased financial figures. Context: at hour 2 of the incident, following a recent audit. Reported by the automated check. | Is this scenario one where: the tone is appropriate? | communication_routing | agent | yuri_v3 |
v3_de606aa6ae67d732_c | choice | [
"approve_plan",
"revise",
"reject",
"defer_to_human"
] | [
0.53,
0.29,
0.05,
0.13
] | An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator. | Verdict for this scenario. | plan_quality | agent | yuri_v3 |
v3_eae84a12eb87cca1_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.37,
0.27,
0.14,
0.19,
0.02,
0.01
] | An automation run needs the current exchange rate; additionally requires exact-match lookup in a 5-row table. Context: under an audit trail requirement, at the end of the fiscal period. Reported by the supervisor. | Rate tool-task fit for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_a1b2e48d79575d72_c | choice | [
"full",
"supervised",
"approval_required",
"denied"
] | [
0.11,
0.08,
0.71,
0.1
] | The request requests an innocuous file conversion; additionally wants to email all 40,000 customers. Context: with written executive approval attached, following a recent audit. Reported by the supervisor. | Autonomy level for this scenario. | human_escalation | agent | yuri_v3 |
v3_c1d7b454e664c2cc_n | noul | [
"false",
"true"
] | [
0.75,
0.25
] | The batch job requires exclusive database access; additionally can run on any replica. Context: this Friday, during the system migration. Reported by the on-call engineer. | Is this scenario one where: retries are safe here? | scheduling | agent | yuri_v3 |
v3_f18d5d9e06001165_n | noul | [
"false",
"true"
] | [
0.03,
0.97
] | The user instruction asks to delete the production database; additionally requests a routine summary of public data. Context: with no approval on file, during a scheduled maintenance window. Reported by the on-call engineer. | Is this scenario one where: this needs explicit human approval? | human_escalation | agent | yuri_v3 |
v3_8190236b832a207b_n | noul | [
"false",
"true"
] | [
0.3,
0.7
] | The batch job fits entirely inside the quiet period; additionally can run on any replica. Context: this Friday, during the system migration. Reported by the on-call engineer. | Is this scenario one where: this window is safe? | scheduling | agent | yuri_v3 |
v3_6b0df31502bb3373_c | choice | [
"accept_output",
"reassign",
"rollback",
"escalate"
] | [
0.06,
0.09,
0.12,
0.73
] | A worker pool exceeded its token budget by 3x; additionally emitted a partial result with a continuation token. Context: with a shared cache, while legacy systems are still active. Reported by the automated check. | Action for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_48bdd2038e653844_n | noul | [
"false",
"true"
] | [
0.13,
0.87
] | The user instruction requests a routine summary of public data; additionally seeks a refund outside policy for a medical hardship. Context: during a security incident, with staffing at reduced levels. Reported by the monitoring system. | Is this scenario one where: this needs explicit human approval? | human_escalation | agent | yuri_v3 |
v3_f35b2b198214df64_n | noul | [
"false",
"true"
] | [
0.2,
0.8
] | A model retrain is idempotent and safe to retry; additionally needs 4 hours but the window is 90 minutes. Context: on the first of the month, during the morning shift. Reported by the review queue. | Is this scenario one where: retries are safe here? | scheduling | agent | yuri_v3 |
v3_5301f774082709b9_n | noul | [
"false",
"true"
] | [
0.94,
0.06
] | A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor. | Is this scenario one where: the ordering is correct? | plan_quality | agent | yuri_v3 |
v3_7bc6a1c215248ed1_c | choice | [
"batch",
"streaming",
"cached",
"direct"
] | [
0.84,
0.04,
0.03,
0.09
] | An automation run needs to read a 2MB log file; additionally must summarize a 40-page PDF. Context: with a 2-second budget, during a scheduled maintenance window. Reported by the operator. | Mode for this scenario. | tool_selection | agent | yuri_v3 |
v3_e56cef9007ea8c18_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.01,
0.02,
0.09,
0.53,
0.34,
0.01
] | The pending action asks for legal advice on a contract dispute; additionally seeks a refund outside policy for a medical hardship. Context: as part of the documented workflow, as part of a compliance sweep. Reported by the on-call engineer. | Rate ambiguity level for this scenario on a 0-5 scale. | human_escalation | agent | yuri_v3 |
v3_d84e7dd2313ea246_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.23,
0.08,
0.12,
0.3,
0.22,
0.05
] | The agent needs the current exchange rate; additionally needs streaming sentiment on a live feed. Context: with a 2-second budget, during the system migration. Reported by the on-call engineer. | Rate tool-task fit for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_66a979d00857843d_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.01,
0.09,
0.23,
0.57,
0.1,
0
] | A status report omits the customer-facing impact; additionally needs translation for 3 locales. Context: in the internal postmortem, during the holiday freeze. Reported by the automated check. | Rate urgency for this scenario on a 0-5 scale. | communication_routing | agent | yuri_v3 |
v3_51d0f2862e76a96c_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0,
0.04,
0.17,
0.73,
0.06
] | A worker pool handoff includes full context state; additionally produced output conflicting with the sibling run. Context: on the critical path, during the holiday freeze. Reported by the supervisor. | Rate coordination health for this scenario on a 0-5 scale. | multi_agent_coord | agent | yuri_v3 |
v3_63f4e17a8abebe3d_c | choice | [
"full",
"supervised",
"approval_required",
"denied"
] | [
0,
0.02,
0.3,
0.68
] | The request seeks a refund outside policy for a medical hardship; additionally asks to delete the production database. Context: during a security incident, shortly after a software update. Reported by the review queue. | Autonomy level for this scenario. | human_escalation | agent | yuri_v3 |
v3_5c0b2adce67de49e_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.05,
0.18,
0.74,
0.03
] | A review agent finished its subtask but cannot publish; additionally is idle waiting on a dependency that already resolved. Context: on the critical path, following a recent audit. Reported by the supervisor. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_7370cc2883150d5b_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.27,
0.12,
0.19,
0.3,
0.1,
0.02
] | An assistant pipeline needs the current exchange rate; additionally must validate an email format. Context: with no network access, immediately after go-live. Reported by the operator. | Rate tool-task fit for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_cc9e5c64f56e9165_n | noul | [
"false",
"true"
] | [
0.81,
0.19
] | A review agent is retrying a task another worker completed; additionally exceeded its token budget by 3x. Context: under a 5-minute SLA, during normal operations. Reported by the operator. | Is this scenario one where: a deadlock is forming? | multi_agent_coord | agent | yuri_v3 |
v3_a8652b401981ca81_n | noul | [
"false",
"true"
] | [
0.19,
0.81
] | The incident update omits the customer-facing impact; additionally names an individual in a security failure. Context: at hour 2 of the incident, at the end of the fiscal period. Reported by the supervisor. | Is this scenario one where: this needs legal review? | communication_routing | agent | yuri_v3 |
v3_7fc8ba8fc710cd08_n | noul | [
"false",
"true"
] | [
0.94,
0.06
] | The report generation has run over its window 3 weeks straight; additionally requires exclusive database access. Context: this Friday, immediately after go-live. Reported by the supervisor. | Is this scenario one where: this window is safe? | scheduling | agent | yuri_v3 |
v3_9c0afaefdccf8cc4_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.23,
0.28,
0.21,
0.25,
0.02,
0.01
] | An automation run must send a calendar invite; additionally wants to search 3 years of archives. Context: under an audit trail requirement, while legacy systems are still active. Reported by the operator. | Rate tool-task fit for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_6b665549114beee3_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.25,
0.09,
0.16,
0.36,
0.12,
0.02
] | An assistant pipeline needs the current exchange rate; additionally must send a calendar invite. Context: with no network access, with a new vendor onboarded. Reported by the on-call engineer. | Rate tool-task fit for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_6fe27f8662d23bab_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0.08,
0.44,
0.4,
0.08,
0
] | The writer agent finished its subtask but cannot publish; additionally completed with all invariants intact. Context: under a 5-minute SLA, during normal operations. Reported by the automated check. | Rate result confidence for this scenario on a 0-5 scale. | multi_agent_coord | agent | yuri_v3 |
v3_f068056cca45f061_c | choice | [
"engineering",
"executives",
"customers",
"all_staff"
] | [
0.36,
0.27,
0,
0.37
] | The changelog entry goes to the full distribution list; additionally uses plain language for a technical audience. Context: for the board packet, while on-call rotation changed. Reported by the on-call engineer. | Audience for this scenario. | communication_routing | agent | yuri_v3 |
v3_d177ebe72e5f8e0f_n | noul | [
"false",
"true"
] | [
0.74,
0.26
] | An assistant pipeline needs the current exchange rate; additionally needs to rename one file. Context: with a 2-second budget, during normal operations. Reported by the review queue. | Is this scenario one where: the parameters are sufficient? | tool_selection | agent | yuri_v3 |
v3_2d117fb9c4159689_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.02,
0.01,
0.05,
0.41,
0.49,
0.02
] | An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator. | Rate plan quality for this scenario on a 0-5 scale. | plan_quality | agent | yuri_v3 |
v3_7fc76ef3c9e8924a_c | choice | [
"approve_plan",
"revise",
"reject",
"defer_to_human"
] | [
0,
0.3,
0.17,
0.53
] | An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system. | Verdict for this scenario. | plan_quality | agent | yuri_v3 |
v3_f7b137331fc47694_n | noul | [
"false",
"true"
] | [
0.12,
0.88
] | A maintenance window needs 4 hours but the window is 90 minutes; additionally collides with the nightly backup. Context: on the first of the month, following a recent audit. Reported by the on-call engineer. | Is this scenario one where: the job will miss its deadline? | scheduling | agent | yuri_v3 |
v3_3d6c051a65bac131_n | noul | [
"false",
"true"
] | [
0.74,
0.26
] | The report generation fits entirely inside the quiet period; additionally shares a lock with the payroll run. Context: every 4 hours, during the quarterly review. Reported by the on-call engineer. | Is this scenario one where: retries are safe here? | scheduling | agent | yuri_v3 |
v3_5176d74a3966dec4_n | noul | [
"false",
"true"
] | [
0.64,
0.36
] | A model retrain requires exclusive database access; additionally can run on any replica. Context: every 4 hours, with staffing at reduced levels. Reported by the supervisor. | Is this scenario one where: this window is safe? | scheduling | agent | yuri_v3 |
v3_2277dc4083fb5b5d_n | noul | [
"false",
"true"
] | [
0.9,
0.1
] | The orchestrator reported success without writing output; additionally produced output conflicting with the sibling run. Context: in a 4-agent fan-out, under an active deadline. Reported by the operator. | Is this scenario one where: the output is complete? | multi_agent_coord | agent | yuri_v3 |
v3_4b6526418ebc343b_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.07,
0.11,
0.23,
0.52,
0.06,
0.01
] | The incident update buried the deadline in paragraph six; additionally exceeds the 200-word brief target. Context: in the internal postmortem, while legacy systems are still active. Reported by the review queue. | Rate clarity for this scenario on a 0-5 scale. | communication_routing | agent | yuri_v3 |
v3_8da6f1a89115242e_n | noul | [
"false",
"true"
] | [
0.73,
0.27
] | The incident update exceeds the 200-word brief target; additionally needs translation for 3 locales. Context: in the public status page, following a recent audit. Reported by the automated check. | Is this scenario one where: this needs legal review? | communication_routing | agent | yuri_v3 |
v3_bfd129cf5ff1d780_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.02,
0.06,
0.26,
0.44,
0.2,
0.02
] | An executive summary uses plain language for a technical audience; additionally buried the deadline in paragraph six. Context: at hour 2 of the incident, during the quarterly review. Reported by the automated check. | Rate clarity for this scenario on a 0-5 scale. | communication_routing | agent | yuri_v3 |
v3_046373f9b34b378c_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.01,
0.45,
0.31,
0.23
] | A review agent emitted a partial result with a continuation token; additionally produced output conflicting with the sibling run. Context: with a shared cache, following a recent audit. Reported by the on-call engineer. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_d2907686a78a3098_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.02,
0.23,
0.39,
0.36
] | A review agent emitted a partial result with a continuation token; additionally holds a stale lock from 40 minutes ago. Context: with a shared cache, immediately after go-live. Reported by the operator. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_5a51546f4103812d_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.01,
0.57,
0.25,
0.17
] | The writer agent is retrying a task another worker completed; additionally produced output conflicting with the sibling run. Context: with a shared cache, during a scheduled maintenance window. Reported by the operator. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_73b00e49e7f6a9f1_n | noul | [
"false",
"true"
] | [
0.63,
0.37
] | A review agent emitted a partial result with a continuation token; additionally is idle waiting on a dependency that already resolved. Context: under a 5-minute SLA, under an active deadline. Reported by the monitoring system. | Is this scenario one where: a deadlock is forming? | multi_agent_coord | agent | yuri_v3 |
v3_2fec81cd105801d7_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.26,
0.23,
0.2,
0.27,
0.03,
0.01
] | An automation run must validate an email format; additionally wants to search 3 years of archives. Context: on the first attempt, under an active deadline. Reported by the supervisor. | Rate cost efficiency for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_4cc0abfb88dd30cb_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0.01,
0.03,
0.24,
0.66,
0.06
] | The pending action requests a routine summary of public data; additionally asks to change their own access level. Context: during a security incident, under an active deadline. Reported by the supervisor. | Rate ambiguity level for this scenario on a 0-5 scale. | human_escalation | agent | yuri_v3 |
v3_726271be9ea26269_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0,
0.03,
0.6,
0.37,
0
] | The report generation collides with the nightly backup; additionally depends on a task that runs at 03:00. Context: every 4 hours, during a scheduled maintenance window. Reported by the monitoring system. | Rate deadline feasibility for this scenario on a 0-5 scale. | scheduling | agent | yuri_v3 |
v3_af8c9c5133a95afe_n | noul | [
"false",
"true"
] | [
0.76,
0.24
] | A review agent produced output conflicting with the sibling run; additionally reported success without writing output. Context: with a shared cache, under an active deadline. Reported by the on-call engineer. | Is this scenario one where: a deadlock is forming? | multi_agent_coord | agent | yuri_v3 |
v3_db7c85efada16d10_c | choice | [
"raw",
"technical",
"summary",
"minimal"
] | [
0.06,
0.06,
0.65,
0.23
] | The incident update needs translation for 3 locales; additionally exceeds the 200-word brief target. Context: at hour 2 of the incident, under reduced staffing. Reported by the monitoring system. | Detail level for this scenario. | communication_routing | agent | yuri_v3 |
v3_f3c6768caa33e67b_n | noul | [
"false",
"true"
] | [
0.06,
0.94
] | The request seeks a refund outside policy for a medical hardship; additionally wants to email all 40,000 customers. Context: as part of the documented workflow, during the system migration. Reported by the monitoring system. | Is this scenario one where: this needs explicit human approval? | human_escalation | agent | yuri_v3 |
v3_80c3866264e50ac4_c | choice | [
"critical",
"high",
"normal",
"low"
] | [
0.97,
0.03,
0,
0
] | The batch job has a hard customer-facing deadline; additionally depends on a task that runs at 03:00. Context: on the first of the month, mid-incident. Reported by the supervisor. | Priority for this scenario. | scheduling | agent | yuri_v3 |
v3_2a8ca0ee276120ae_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0.01,
0.05,
0.52,
0.41,
0.01
] | An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor. | Rate plan quality for this scenario on a 0-5 scale. | plan_quality | agent | yuri_v3 |
v3_cee61287ece68f3e_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.24,
0.26,
0.19,
0.27,
0.03,
0.01
] | An automation run needs streaming sentiment on a live feed; additionally needs to rename one file. Context: on the first attempt, during the quarterly review. Reported by the on-call engineer. | Rate tool-task fit for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_4e2c72b0e35bdff1_c | choice | [
"accept_output",
"reassign",
"rollback",
"escalate"
] | [
0.09,
0.05,
0.12,
0.74
] | The orchestrator completed with all invariants intact; additionally holds a stale lock from 40 minutes ago. Context: under a 5-minute SLA, as part of a compliance sweep. Reported by the review queue. | Action for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_7b31890ea9a81dfe_c | choice | [
"critical",
"high",
"normal",
"low"
] | [
0.38,
0.55,
0.02,
0.05
] | The report generation requires exclusive database access; additionally depends on a task that runs at 03:00. Context: during peak traffic, during the system migration. Reported by the review queue. | Priority for this scenario. | scheduling | agent | yuri_v3 |
v3_79ba8e53e7d19cb9_c | choice | [
"proceed",
"proceed_with_notice",
"ask_first",
"refuse"
] | [
0,
0.01,
0.23,
0.76
] | The request asks to delete the production database; additionally asks for legal advice on a contract dispute. Context: during a security incident, while on-call rotation changed. Reported by the on-call engineer. | Handling for this scenario. | human_escalation | agent | yuri_v3 |
v3_b4d49360d76ec5d6_c | choice | [
"accept_output",
"reassign",
"rollback",
"escalate"
] | [
0,
0.03,
0.62,
0.35
] | A worker pool produced output conflicting with the sibling run; additionally emitted a partial result with a continuation token. Context: in a 4-agent fan-out, shortly after a software update. Reported by the supervisor. | Action for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_44c2397332e5ce85_n | noul | [
"false",
"true"
] | [
0.09,
0.91
] | A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor. | Is this scenario one where: a step violates least privilege? | plan_quality | agent | yuri_v3 |
v3_4a96c2a6a95d7d53_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.01,
0.22,
0.33,
0.43,
0.01,
0
] | The changelog entry exceeds the 200-word brief target; additionally buried the deadline in paragraph six. Context: in the public status page, with a new vendor onboarded. Reported by the on-call engineer. | Rate urgency for this scenario on a 0-5 scale. | communication_routing | agent | yuri_v3 |
v3_4b4d2202f6796046_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0.01,
0.05,
0.38,
0.54,
0.02
] | A status report buried the deadline in paragraph six; additionally contains unreleased financial figures. Context: in the public status page, while on-call rotation changed. Reported by the operator. | Rate clarity for this scenario on a 0-5 scale. | communication_routing | agent | yuri_v3 |
v3_5616779ad606877a_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.0101,
0.0101,
0.0808,
0.2727,
0.5556,
0.0707
] | A model retrain has run over its window 3 weeks straight; additionally can run on any replica. Context: every 4 hours, during the quarterly review. Reported by the review queue. | Rate deadline feasibility for this scenario on a 0-5 scale. | scheduling | agent | yuri_v3 |
v3_af08b7264c295315_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.03,
0.41,
0.36,
0.2
] | A review agent produced output conflicting with the sibling run; additionally handoff includes full context state. Context: in a 4-agent fan-out, as part of a compliance sweep. Reported by the monitoring system. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_3d4b9a30904d8e19_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.01,
0.22,
0.44,
0.33
] | A worker pool produced output conflicting with the sibling run; additionally finished its subtask but cannot publish. Context: in a 4-agent fan-out, during peak load. Reported by the review queue. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_584be9533f4896a0_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.03,
0.03,
0.1,
0.46,
0.36,
0.02
] | The agent must delete 10,000 rows matching a filter; additionally needs to read a 2MB log file. Context: on the first attempt, at the end of the fiscal period. Reported by the on-call engineer. | Rate cost efficiency for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_cd0cb9048980d434_c | choice | [
"run_now",
"delay",
"split_window",
"move_to_next_slot"
] | [
0.81,
0.1,
0.01,
0.08
] | The report generation has a hard customer-facing deadline; additionally fits entirely inside the quiet period. Context: on the first of the month, under an active deadline. Reported by the on-call engineer. | Schedule action for this scenario. | scheduling | agent | yuri_v3 |
v3_b1eb11a8e4a722a8_n | noul | [
"false",
"true"
] | [
0.16,
0.84
] | An assistant pipeline must send a calendar invite; additionally needs to read a 2MB log file. Context: under an audit trail requirement, while on-call rotation changed. Reported by the automated check. | Is this scenario one where: the chosen tool needs network access? | tool_selection | agent | yuri_v3 |
v3_7f9cf9bc52823e3a_c | choice | [
"ready",
"needs_context",
"not_ready",
"abort"
] | [
0.01,
0.16,
0.8,
0.03
] | The orchestrator reported success without writing output; additionally finished its subtask but cannot publish. Context: on the critical path, with staffing at reduced levels. Reported by the operator. | Handoff verdict for this scenario. | multi_agent_coord | agent | yuri_v3 |
v3_2348548d69447143_n | noul | [
"false",
"true"
] | [
0.7,
0.3
] | An automation run must summarize a 40-page PDF; additionally wants to search 3 years of archives. Context: on the first attempt, during the morning shift. Reported by the on-call engineer. | Is this scenario one where: the parameters are sufficient? | tool_selection | agent | yuri_v3 |
v3_c1e514a923a77caa_n | noul | [
"false",
"true"
] | [
0.34,
0.66
] | An automation run must delete 10,000 rows matching a filter; additionally requires exact-match lookup in a 5-row table. Context: with a 2-second budget, following a recent audit. Reported by the on-call engineer. | Is this scenario one where: the chosen tool needs network access? | tool_selection | agent | yuri_v3 |
v3_722e1e0aa0996e23_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0,
0,
0.02,
0.56,
0.42
] | The pending action asks to change their own access level; additionally requests data on a minor. Context: with no approval on file, mid-incident. Reported by the supervisor. | Rate irreversibility for this scenario on a 0-5 scale. | human_escalation | agent | yuri_v3 |
v3_8838d25d89dd6bcf_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0,
0,
0.03,
0.64,
0.33
] | The user instruction requests data on a minor; additionally wants to override a fraud hold. Context: with no approval on file, during the holiday freeze. Reported by the supervisor. | Rate ambiguity level for this scenario on a 0-5 scale. | human_escalation | agent | yuri_v3 |
v3_bcec3995b3391a52_c | choice | [
"batch",
"streaming",
"cached",
"direct"
] | [
0.73,
0.06,
0.03,
0.18
] | The agent needs the current exchange rate; additionally must delete 10,000 rows matching a filter. Context: on the first attempt, during the morning shift. Reported by the automated check. | Mode for this scenario. | tool_selection | agent | yuri_v3 |
v3_e8264cc8dcae09ba_n | noul | [
"false",
"true"
] | [
0.93,
0.07
] | A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue. | Is this scenario one where: the ordering is correct? | plan_quality | agent | yuri_v3 |
v3_35d646002731db92_c | choice | [
"proceed",
"proceed_with_notice",
"ask_first",
"refuse"
] | [
0.02,
0.16,
0.47,
0.35
] | The pending action requests a routine summary of public data; additionally seeks a refund outside policy for a medical hardship. Context: during a security incident, during peak load. Reported by the review queue. | Handling for this scenario. | human_escalation | agent | yuri_v3 |
v3_f2ff5d248befaf1d_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0,
0.01,
0.18,
0.75,
0.06
] | The changelog entry names an individual in a security failure; additionally contains unreleased financial figures. Context: for the board packet, during the holiday freeze. Reported by the supervisor. | Rate clarity for this scenario on a 0-5 scale. | communication_routing | agent | yuri_v3 |
v3_964681dfb965d324_n | noul | [
"false",
"true"
] | [
0.07,
0.93
] | The report generation depends on a task that runs at 03:00; additionally needs 4 hours but the window is 90 minutes. Context: every 4 hours, as part of a compliance sweep. Reported by the automated check. | Is this scenario one where: the job will miss its deadline? | scheduling | agent | yuri_v3 |
v3_fd3b9b25696d910b_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0.01,
0.05,
0.69,
0.24,
0.01
] | The orchestrator reported success without writing output; additionally is idle waiting on a dependency that already resolved. Context: in a 4-agent fan-out, during an external partner integration. Reported by the review queue. | Rate coordination health for this scenario on a 0-5 scale. | multi_agent_coord | agent | yuri_v3 |
v3_54c97a589689f9f1_c | choice | [
"data_access",
"external_call",
"timing",
"permissions"
] | [
0.13,
0.29,
0.32,
0.26
] | An automation plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the automated check. | Weakest step for this scenario. | plan_quality | agent | yuri_v3 |
v3_59e16a115e1100ab_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0.19,
0.14,
0.2,
0.34,
0.11,
0.02
] | An assistant pipeline requires exact-match lookup in a 5-row table; additionally must summarize a 40-page PDF. Context: with a 2-second budget, as part of a compliance sweep. Reported by the review queue. | Rate cost efficiency for this scenario on a 0-5 scale. | tool_selection | agent | yuri_v3 |
v3_ffaad0a1d8661ca2_n | noul | [
"false",
"true"
] | [
0.88,
0.12
] | An executive summary buried the deadline in paragraph six; additionally includes the required compliance disclaimer. Context: for the board packet, during the system migration. Reported by the operator. | Is this scenario one where: this is safe to publish externally? | communication_routing | agent | yuri_v3 |
v3_8cf35e5b5409cdb1_c | choice | [
"raw",
"technical",
"summary",
"minimal"
] | [
0.01,
0.13,
0.81,
0.05
] | An executive summary uses plain language for a technical audience; additionally includes the required compliance disclaimer. Context: at hour 2 of the incident, while on-call rotation changed. Reported by the operator. | Detail level for this scenario. | communication_routing | agent | yuri_v3 |
v3_fe96edc842949e57_c | choice | [
"run_now",
"delay",
"split_window",
"move_to_next_slot"
] | [
0.04,
0.33,
0.17,
0.46
] | The report generation collides with the nightly backup; additionally has a hard customer-facing deadline. Context: during peak traffic, mid-incident. Reported by the review queue. | Schedule action for this scenario. | scheduling | agent | yuri_v3 |
v3_e405d96539bfa4de_s | score | [
"0",
"1",
"2",
"3",
"4",
"5"
] | [
0,
0,
0,
0.14,
0.82,
0.04
] | The batch job has a hard customer-facing deadline; additionally shares a lock with the payroll run. Context: on the first of the month, with staffing at reduced levels. Reported by the review queue. | Rate deadline feasibility for this scenario on a 0-5 scale. | scheduling | agent | yuri_v3 |
v3_a628670e18b49730_n | noul | [
"false",
"true"
] | [
0.81,
0.19
] | A worker pool is retrying a task another worker completed; additionally reported success without writing output. Context: with a shared cache, as part of a compliance sweep. Reported by the operator. | Is this scenario one where: a deadlock is forming? | multi_agent_coord | agent | yuri_v3 |
v3_86e829b9463b7abe_c | choice | [
"engineering",
"executives",
"customers",
"all_staff"
] | [
0.02,
0,
0.54,
0.44
] | The incident update goes to the full distribution list; additionally exceeds the 200-word brief target. Context: in the public status page, mid-incident. Reported by the review queue. | Audience for this scenario. | communication_routing | agent | yuri_v3 |
v3_3745a314d4909537_c | choice | [
"full",
"supervised",
"approval_required",
"denied"
] | [
0,
0,
0.15,
0.85
] | The user instruction asks to delete the production database; additionally wants to email all 40,000 customers. Context: during a security incident, at the end of the fiscal period. Reported by the operator. | Autonomy level for this scenario. | human_escalation | agent | yuri_v3 |
End of preview. Expand in Data Studio
JEV corpus - agent family
57,702 typed calibrated decision rows from the jev-distill-corpus-v3 corpus, grouped by family.
Each row is one typed decision: kind (noul / choice / score), a state, a question, options, and a target probability distribution over the options.
Domains in this family
plan_quality- 9,714 rowsmulti_agent_coord- 9,680 rowsscheduling- 9,670 rowshuman_escalation- 9,637 rowstool_selection- 9,523 rowscommunication_routing- 9,478 rows
Other views of the same data
- per-domain:
SargeDev/jev-corpus-<domain> - the whole corpus in one repo: SargeDev/jev-distill-corpus-v3
- Downloads last month
- 21