Datasets:
Formats:
csv
Languages:
English
Size:
< 1K
Tags:
quantization
llm-evaluation
model-compression
agent-evaluation
tool-calling
behavioral-evaluation
License:
run_id stringclasses 1
value | timestamp stringlengths 15 15 | runner stringclasses 1
value | family stringclasses 8
values | case_id stringlengths 6 11 | mode stringclasses 4
values | secs float64 0.02 3.21 | timeout int64 0 0 | output_raw stringlengths 1 505 ⌀ | error float64 | plan_len float64 1 6 ⌀ | plan_exact_match float64 0 1 ⌀ | plan_prefix_match_len float64 1 6 ⌀ | plan_stop_idx_expected float64 -1 4 ⌀ | plan_stop_idx_got float64 -1 4 ⌀ | final_equiv_ok float64 0 1 ⌀ | final_constraints_ok float64 0 1 ⌀ | bucket_score int64 0 11 | detail stringclasses 5
values | expected_plan stringlengths 1 40 ⌀ | got_plan stringlengths 1 40 ⌀ | expected_final stringlengths 65 65 ⌀ | got_final stringlengths 65 65 ⌀ | answer_line_ok float64 1 1 ⌀ | args_ok float64 1 1 ⌀ | best_of_k_candidate_idx float64 0 0 ⌀ | best_of_k_k float64 3 3 ⌀ | best_of_k_runner stringclasses 1
value | best_of_k_seed float64 1.34k 1.74k ⌀ | best_of_k_selected_idx float64 0 0 ⌀ | best_of_k_selected_sim float64 0 0 ⌀ | best_of_k_shot1_tier1_pass float64 0 1 ⌀ | best_of_k_tier1_pass_at_k float64 0 1 ⌀ | best_of_k_used float64 1 1 ⌀ | cache_key stringlengths 26 72 | checks_consistent_ok float64 0 1 ⌀ | checks_ok float64 0 1 ⌀ | choice_extracted stringclasses 3
values | constraints_detail stringclasses 1
value | detail_stage1 stringclasses 1
value | expected_answer_line stringclasses 2
values | expected_args stringclasses 2
values | expected_json stringclasses 2
values | expected_turn1 stringclasses 2
values | expected_turn2 stringclasses 2
values | final_consistent_ok float64 0 0 ⌀ | final_match_reported float64 0 0 ⌀ | got_answer_line stringclasses 2
values | got_args stringclasses 2
values | got_json stringclasses 2
values | got_turn1 stringclasses 2
values | got_turn2 stringclasses 2
values | json_exact_match float64 1 1 ⌀ | json_parse_ok float64 1 1 ⌀ | oracle_equiv_ok float64 0 1 ⌀ | oracle_trace stringlengths 209 828 ⌀ | parse_ok float64 1 1 ⌀ | plan_len_ok float64 0 1 ⌀ | schema_errors stringclasses 2
values | schema_ok float64 0 1 ⌀ | stage1_tool_parse_ok float64 1 1 ⌀ | stage1_tool_schema_ok float64 1 1 ⌀ | stop_semantics_ok float64 0 1 ⌀ | tool_exec_detail stringclasses 1
value | tool_exec_ok float64 1 1 ⌀ | tool_name_ok float64 1 1 ⌀ | tool_output_json stringclasses 2
values | tool_payload_used stringclasses 2
values | turn1_exact_match float64 1 1 ⌀ | turn1_parse_ok float64 1 1 ⌀ | turn2_exact_match float64 1 1 ⌀ | turn2_parse_ok float64 1 1 ⌀ |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184753 | quantized_vllm | json | json_01 | tool_decisions_compressed_tensors_w4a16_asym | 1.21 | 0 | {
"tool_name": "place_item",
"args": {
"choice": "A"
}
}
{
"tool_name": "place_item",
"args": {
"choice": "A"
}
}
{
"tool_name": "place_item",
"args": {
"choice": "C"
}
} | null | 3 | 1 | 3 | -1 | -1 | 1 | 1 | 10 | ok | ["A","A","C"] | ["A","A","C"] | {"A":[3,3,0],"B":[2,0,0],"C":[0,0,2],"new_shelf":0,"violation":1} | {"A":[3,3,0],"B":[2,0,0],"C":[0,0,2],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|json|json_01|tool_decisions_compressed_tensors_w4a16_asym | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[2, 2, 0] B=[2, 0, 0] C=[0, 0, 1]
Arrivals=['P', 'S', 'H']
Arrival 1: P
Try A: can_place=1 state=[2, 2, 0]
Placed on A: new_state=[2, 3, 0]
Arrival 2: S
Try A: can_place=1 state=[2, 3, 0]
Placed on A: new_state=[3, 3, 0]
Arrival 3: H
Try A: can_place=0 state=[3, 3, 0]
Try B: can_place=0 state=[2, ... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184753 | quantized_vllm | json | json_02 | tool_decisions_compressed_tensors_w4a16_asym | 1.26 | 0 | {
"tool_name": "place_item",
"args": {
"choice": "A"
}
}
{
"tool_name": "place_item",
"args": {
"choice": "A"
}
}
{
"tool_name": "place_item",
"args": {
"choice": "A"
}
} | null | 3 | 1 | 3 | -1 | -1 | 1 | 1 | 10 | ok | ["A","A","A"] | ["A","A","A"] | {"A":[3,1,0],"B":[0,1,0],"C":[0,0,1],"new_shelf":0,"violation":1} | {"A":[3,1,0],"B":[0,1,0],"C":[0,0,1],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|json|json_02|tool_decisions_compressed_tensors_w4a16_asym | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[1, 0, 0] B=[0, 1, 0] C=[0, 0, 1]
Arrivals=['S', 'S', 'P']
Arrival 1: S
Try A: can_place=1 state=[1, 0, 0]
Placed on A: new_state=[2, 0, 0]
Arrival 2: S
Try A: can_place=1 state=[2, 0, 0]
Placed on A: new_state=[3, 0, 0]
Arrival 3: P
Try A: can_place=1 state=[3, 0, 0]
Placed on A: new_state=[3, 1,... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184753 | quantized_vllm | json | json_03 | tool_decisions_compressed_tensors_w4a16_asym | 1.23 | 0 | {
"tool_name": "place_item",
"args": {
"choice": "C"
}
}
{
"tool_name": "place_item",
"args": {
"choice": "C"
}
}
{
"tool_name": "place_item",
"args": {
"choice": "A"
}
} | null | 3 | 1 | 3 | -1 | -1 | 1 | 1 | 10 | ok | ["C","C","A"] | ["C","C","A"] | {"A":[3,1,0],"B":[0,3,0],"C":[0,0,4],"new_shelf":0,"violation":1} | {"A":[3,1,0],"B":[0,3,0],"C":[0,0,4],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|json|json_03|tool_decisions_compressed_tensors_w4a16_asym | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[3, 0, 0] B=[0, 3, 0] C=[0, 0, 2]
Arrivals=['H', 'H', 'P']
Arrival 1: H
Try A: can_place=0 state=[3, 0, 0]
Try B: can_place=0 state=[0, 3, 0]
Try C: can_place=1 state=[0, 0, 2]
Placed on C: new_state=[0, 0, 3]
Arrival 2: H
Try A: can_place=0 state=[3, 0, 0]
Try B: can_place=0 state=[0, 3, 0]
Try... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184753 | quantized_vllm | json | json_04 | tool_decisions_compressed_tensors_w4a16_asym | 1.11 | 0 | {"tool_name": "place_item", "args": {"choice": "A"}}
{
"tool_name": "place_item",
"args": {
"choice": "A"
}
}
{
"tool_name": "place_item",
"args": {
"choice": "B"
}
} | null | 3 | 1 | 3 | -1 | -1 | 1 | 1 | 10 | ok | ["A","A","B"] | ["A","A","B"] | {"A":[1,1,0],"B":[0,0,1],"C":[0,0,0],"new_shelf":0,"violation":1} | {"A":[1,1,0],"B":[0,0,1],"C":[0,0,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|json|json_04|tool_decisions_compressed_tensors_w4a16_asym | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 0, 0] B=[0, 0, 0] C=[0, 0, 0]
Arrivals=['S', 'P', 'H']
Arrival 1: S
Try A: can_place=1 state=[0, 0, 0]
Placed on A: new_state=[1, 0, 0]
Arrival 2: P
Try A: can_place=1 state=[1, 0, 0]
Placed on A: new_state=[1, 1, 0]
Arrival 3: H
Try A: can_place=0 state=[1, 1, 0]
Try B: can_place=1 state=[0, ... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184830 | quantized_vllm | json_multistep | ms_easy_01 | best_of_k | 2.03014 | 0 | {
"plan": ["A"],
"checks": [
{
"step": 1,
"item": "P",
"choice": "A",
"can_place": 1
}
],
"final": {
"A": [2, 2, 1],
"B": [3, 0, 0],
"C": [0, 0, 1],
"new_shelf": 0,
"violation": 0
},
"stop": false,
"stop_reason": "",
"stop_step_index": -1
} | null | 1 | 1 | 1 | -1 | -1 | 1 | 1 | 1 | ok | ["A"] | ["A"] | {"A":[2,3,0],"B":[3,0,0],"C":[0,0,1],"new_shelf":0,"violation":1} | {"A":[2,3,0],"B":[3,0,0],"C":[0,0,1],"new_shelf":0,"violation":1} | null | null | 0 | 3 | quantized_vllm | 1,337 | 0 | 0 | 1 | 1 | 1 | quantized_vllm|json_multistep|ms_easy_01|best_of_k | 1 | 1 | null | null | null | null | null | null | null | null | 0 | 0 | null | null | null | null | null | null | null | 1 | Initial A=[2, 2, 0] B=[3, 0, 0] C=[0, 0, 1]
Arrivals=['P']
Arrival 1: P
Try A: can_place=1 state=[2, 2, 0]
Placed on A: new_state=[2, 3, 0]
Final A=[2, 3, 0] B=[3, 0, 0] C=[0, 0, 1] new_shelf=0 violation=1 | 1 | 1 | null | 1 | null | null | 1 | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184830 | quantized_vllm | json_multistep | ms_easy_02 | best_of_k | 2.202546 | 0 | {
"plan": ["A", "STOP"],
"checks": [
{
"step": 1,
"item": "S",
"choice": "A",
"can_place": 1
}
],
"final": {
"A": [3, 0, 1],
"B": [0, 0, 0],
"C": [0, 0, 0],
"new_shelf": 1,
"violation": 0
},
"stop": true,
"stop_reason": "No more items to place",
"stop_... | null | 2 | 0 | 1 | -1 | 1 | 0 | 0 | 1 | checks_consistent_ok+oracle_equiv_ok | ["A","A"] | ["A","STOP"] | {"A":[4,1,0],"B":[0,0,0],"C":[0,0,0],"new_shelf":0,"violation":1} | {"A":[4,0,0],"B":[0,0,0],"C":[0,0,0],"new_shelf":1,"violation":1} | null | null | 0 | 3 | quantized_vllm | 1,437 | 0 | 0 | 0 | 0 | 1 | quantized_vllm|json_multistep|ms_easy_02|best_of_k | 0 | 0 | null | null | null | null | null | null | null | null | 0 | 0 | null | null | null | null | null | null | null | 0 | Initial A=[3, 0, 0] B=[0, 0, 0] C=[0, 0, 0]
Arrivals=['S', 'P']
Arrival 1: S
Try A: can_place=1 state=[3, 0, 0]
Placed on A: new_state=[4, 0, 0]
Arrival 2: P
Try A: can_place=1 state=[4, 0, 0]
Placed on A: new_state=[4, 1, 0]
Final A=[4, 1, 0] B=[0, 0, 0] C=[0, 0, 0] new_shelf=0 violation=1 | 1 | 1 | null | 1 | null | null | 1 | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184830 | quantized_vllm | json_multistep | ms_med_01 | best_of_k | 2.754004 | 0 | {
"plan": ["A", "B", "C"],
"checks": [
{"step": 1, "item": "S", "choice": "A", "can_place": 1},
{"step": 2, "item": "P", "choice": "B", "can_place": 1},
{"step": 3, "item": "H", "choice": "C", "can_place": 1}
],
"final": {
"A": [2, 3, 0],
"B": [0, 0, 0],
"C": [0, 0, 2],
"new_shelf": ... | null | 3 | 1 | 3 | -1 | -1 | 1 | 1 | 3 | ok | ["A","B","C"] | ["A","B","C"] | {"A":[3,3,0],"B":[0,1,0],"C":[0,0,3],"new_shelf":0,"violation":1} | {"A":[3,3,0],"B":[0,1,0],"C":[0,0,3],"new_shelf":0,"violation":1} | null | null | 0 | 3 | quantized_vllm | 1,537 | 0 | 0 | 1 | 1 | 1 | quantized_vllm|json_multistep|ms_med_01|best_of_k | 1 | 1 | null | null | null | null | null | null | null | null | 0 | 0 | null | null | null | null | null | null | null | 1 | Initial A=[2, 3, 0] B=[0, 0, 0] C=[0, 0, 2]
Arrivals=['S', 'P', 'H']
Arrival 1: S
Try A: can_place=1 state=[2, 3, 0]
Placed on A: new_state=[3, 3, 0]
Arrival 2: P
Try A: can_place=0 state=[3, 3, 0]
Try B: can_place=1 state=[0, 0, 0]
Placed on B: new_state=[0, 1, 0]
Arrival 3: H
Try A: can_place=0 state=[3, ... | 1 | 1 | null | 1 | null | null | 1 | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184830 | quantized_vllm | json_multistep | ms_med_02 | best_of_k | 3.213176 | 0 | {
"plan": ["A", "B", "C", "STOP"],
"checks": [
{"step": 1, "item": "P", "choice": "A", "can_place": 1},
{"step": 2, "item": "S", "choice": "B", "can_place": 1},
{"step": 3, "item": "H", "choice": "C", "can_place": 1},
{"step": 4, "item": "STOP", "choice": "STOP", "can_place": 0}
],
"final": {
... | null | 4 | 0 | 3 | -1 | -1 | 1 | 0 | 3 | checks_consistent_ok+stop_semantics_ok | ["A","B","C"] | ["A","B","C"] | {"A":[5,1,0],"B":[1,4,0],"C":[0,0,2],"new_shelf":0,"violation":1} | {"A":[5,1,0],"B":[1,4,0],"C":[0,0,2],"new_shelf":0,"violation":1} | null | null | 0 | 3 | quantized_vllm | 1,637 | 0 | 0 | 0 | 0 | 1 | quantized_vllm|json_multistep|ms_med_02|best_of_k | 0 | 0 | null | null | null | null | null | null | null | null | 0 | 0 | null | null | null | null | null | null | null | 1 | Initial A=[5, 0, 0] B=[0, 4, 0] C=[0, 0, 1]
Arrivals=['P', 'S', 'H']
Arrival 1: P
Try A: can_place=1 state=[5, 0, 0]
Placed on A: new_state=[5, 1, 0]
Arrival 2: S
Try A: can_place=0 state=[5, 1, 0]
Try B: can_place=1 state=[0, 4, 0]
Placed on B: new_state=[1, 4, 0]
Arrival 3: H
Try A: can_place=0 state=[5, ... | 1 | 0 | null | 1 | null | null | 0 | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184830 | quantized_vllm | json_multistep | ms_hard_01 | best_of_k | 2.814817 | 0 | {
"plan": ["A", "A", "STOP"],
"checks": [
{"step": 1, "item": "S", "choice": "A", "can_place": 1},
{"step": 2, "item": "S", "choice": "A", "can_place": 1},
{"step": 3, "item": "H", "choice": "STOP", "can_place": 0}
],
"final": {
"A": [1, 5, 0],
"B": [2, 0, 0],
"C": [0, 0, 0],
"new_sh... | null | 3 | 0 | 1 | -1 | 2 | 0 | 0 | 1 | checks_consistent_ok+oracle_equiv_ok | ["A","B","C"] | ["A","A","STOP"] | {"A":[1,5,0],"B":[3,0,0],"C":[0,0,1],"new_shelf":0,"violation":1} | {"A":[1,5,0],"B":[2,0,0],"C":[0,0,0],"new_shelf":1,"violation":1} | null | null | 0 | 3 | quantized_vllm | 1,737 | 0 | 0 | 0 | 0 | 1 | quantized_vllm|json_multistep|ms_hard_01|best_of_k | 0 | 0 | null | null | null | null | null | null | null | null | 0 | 0 | null | null | null | null | null | null | null | 0 | Initial A=[0, 5, 0] B=[2, 0, 0] C=[0, 0, 0]
Arrivals=['S', 'S', 'H']
Arrival 1: S
Try A: can_place=1 state=[0, 5, 0]
Placed on A: new_state=[1, 5, 0]
Arrival 2: S
Try A: can_place=0 state=[1, 5, 0]
Try B: can_place=1 state=[2, 0, 0]
Placed on B: new_state=[3, 0, 0]
Arrival 3: H
Try A: can_place=0 state=[1, ... | 1 | 1 | null | 1 | null | null | 1 | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184831 | quantized_vllm | stateful_followup | state_01 | 2turn | 0.2 | 0 | {"counter": 2}
{"counter": 5} | null | null | null | null | null | null | null | null | 2 | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|stateful_followup|state_01|2turn | null | null | null | null | null | null | null | null | {"counter":2} | {"counter":5} | null | null | null | null | null | {"counter":2} | {"counter":5} | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | 1 | 1 | 1 | 1 |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184831 | quantized_vllm | stateful_followup | state_02 | 2turn | 0.38 | 0 | {"items": ["a", "b"]}
{"items": ["a", "b", "c"]} | null | null | null | null | null | null | null | null | 2 | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|stateful_followup|state_02|2turn | null | null | null | null | null | null | null | null | {"items":["a","b"]} | {"items":["a","b","c"]} | null | null | null | null | null | {"items":["a","b"]} | {"items":["a","b","c"]} | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | 1 | 1 | 1 | 1 |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184832 | quantized_vllm | mixed_brief_json | mixed_01 | null | 0.39 | 0 | ANSWER: 13
{"a": 4, "b": 9, "sum": 13} | null | null | null | null | null | null | null | null | 2 | ok | null | null | null | null | 1 | null | null | null | null | null | null | null | null | null | null | quantized_vllm|mixed_brief_json|mixed_01| | null | null | null | null | null | ANSWER: 13 | null | {"a":4,"b":9,"sum":13} | null | null | null | null | ANSWER: 13 | null | {"a":4,"b":9,"sum":13} | null | null | 1 | 1 | null | null | null | null | null | 1 | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184832 | quantized_vllm | mixed_brief_json | mixed_02 | null | 0.35 | 0 | ANSWER: 6
{"a": -2, "b": 8, "sum": 6} | null | null | null | null | null | null | null | null | 2 | ok | null | null | null | null | 1 | null | null | null | null | null | null | null | null | null | null | quantized_vllm|mixed_brief_json|mixed_02| | null | null | null | null | null | ANSWER: 6 | null | {"a":-2,"b":8,"sum":6} | null | null | null | null | ANSWER: 6 | null | {"a":-2,"b":8,"sum":6} | null | null | 1 | 1 | null | null | null | null | null | 1 | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184834 | quantized_vllm | fuzz | fuzz_0000 | tool_decisions | 2.05 | 0 | null | null | 5 | 1 | 5 | -1 | -1 | 1 | 1 | 10 | ok | ["A","B","B","B","B"] | ["A","B","B","B","B"] | {"A":[0,6,0],"B":[5,1,0],"C":[5,0,0],"new_shelf":0,"violation":1} | {"A":[0,6,0],"B":[5,1,0],"C":[5,0,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0000|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 5, 0] B=[1, 1, 0] C=[5, 0, 0]
Arrivals=['P', 'S', 'S', 'S', 'S']
Arrival 1: P
Try A: can_place=1 state=[0, 5, 0]
Placed on A: new_state=[0, 6, 0]
Arrival 2: S
Try A: can_place=0 state=[0, 6, 0]
Try B: can_place=1 state=[1, 1, 0]
Placed on B: new_state=[2, 1, 0]
Arrival 3: S
Try A: can_place=0 ... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184834 | quantized_vllm | fuzz | fuzz_0001 | tool_decisions | 0.42 | 0 | null | null | 1 | 1 | 1 | -1 | -1 | 1 | 1 | 10 | ok | ["A"] | ["A"] | {"A":[5,0,0],"B":[5,1,0],"C":[4,1,0],"new_shelf":0,"violation":1} | {"A":[5,0,0],"B":[5,1,0],"C":[4,1,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0001|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[4, 0, 0] B=[5, 1, 0] C=[4, 1, 0]
Arrivals=['S']
Arrival 1: S
Try A: can_place=1 state=[4, 0, 0]
Placed on A: new_state=[5, 0, 0]
Final A=[5, 0, 0] B=[5, 1, 0] C=[4, 1, 0] new_shelf=0 violation=1 | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184835 | quantized_vllm | fuzz | fuzz_0002 | tool_decisions | 0.41 | 0 | null | null | 1 | 1 | 1 | -1 | -1 | 1 | 1 | 10 | ok | ["A"] | ["A"] | {"A":[2,3,0],"B":[0,0,6],"C":[0,3,0],"new_shelf":0,"violation":1} | {"A":[2,3,0],"B":[0,0,6],"C":[0,3,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0002|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[2, 2, 0] B=[0, 0, 6] C=[0, 3, 0]
Arrivals=['P']
Arrival 1: P
Try A: can_place=1 state=[2, 2, 0]
Placed on A: new_state=[2, 3, 0]
Final A=[2, 3, 0] B=[0, 0, 6] C=[0, 3, 0] new_shelf=0 violation=1 | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184835 | quantized_vllm | fuzz | fuzz_0003 | tool_decisions | 0.75 | 0 | null | null | 5 | 1 | 5 | 1 | 1 | 1 | 1 | 10 | ok | ["A","STOP","STOP","STOP","STOP"] | ["A","STOP","STOP","STOP","STOP"] | {"A":[4,2,0],"B":[5,1,0],"C":[3,0,0],"new_shelf":1,"violation":1} | {"A":[4,2,0],"B":[5,1,0],"C":[3,0,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0003|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[4, 1, 0] B=[5, 1, 0] C=[3, 0, 0]
Arrivals=['P', 'H', 'H', 'P', 'H']
Arrival 1: P
Try A: can_place=1 state=[4, 1, 0]
Placed on A: new_state=[4, 2, 0]
Arrival 2: H
Try A: can_place=0 state=[4, 2, 0]
Try B: can_place=0 state=[5, 1, 0]
Try C: can_place=0 state=[3, 0, 0]
No placement possible on A/B/C... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184836 | quantized_vllm | fuzz | fuzz_0004 | tool_decisions | 0.82 | 0 | null | null | 2 | 1 | 2 | -1 | -1 | 1 | 1 | 10 | ok | ["A","A"] | ["A","A"] | {"A":[1,1,0],"B":[0,0,1],"C":[0,6,0],"new_shelf":0,"violation":1} | {"A":[1,1,0],"B":[0,0,1],"C":[0,6,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0004|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 0, 0] B=[0, 0, 1] C=[0, 6, 0]
Arrivals=['S', 'P']
Arrival 1: S
Try A: can_place=1 state=[0, 0, 0]
Placed on A: new_state=[1, 0, 0]
Arrival 2: P
Try A: can_place=1 state=[1, 0, 0]
Placed on A: new_state=[1, 1, 0]
Final A=[1, 1, 0] B=[0, 0, 1] C=[0, 6, 0] new_shelf=0 violation=1 | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184838 | quantized_vllm | fuzz | fuzz_0005 | tool_decisions | 2.03 | 0 | null | null | 5 | 1 | 5 | -1 | -1 | 1 | 1 | 10 | ok | ["B","C","B","B","C"] | ["B","C","B","B","C"] | {"A":[5,1,0],"B":[0,0,5],"C":[4,0,0],"new_shelf":0,"violation":1} | {"A":[5,1,0],"B":[0,0,5],"C":[4,0,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0005|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[5, 1, 0] B=[0, 0, 2] C=[2, 0, 0]
Arrivals=['H', 'S', 'H', 'H', 'S']
Arrival 1: H
Try A: can_place=0 state=[5, 1, 0]
Try B: can_place=1 state=[0, 0, 2]
Placed on B: new_state=[0, 0, 3]
Arrival 2: S
Try A: can_place=0 state=[5, 1, 0]
Try B: can_place=0 state=[0, 0, 3]
Try C: can_place=1 state=[2, 0... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184839 | quantized_vllm | fuzz | fuzz_0006 | tool_decisions | 0.71 | 0 | null | null | 2 | 1 | 2 | 1 | 1 | 1 | 1 | 10 | ok | ["B","STOP"] | ["B","STOP"] | {"A":[0,0,3],"B":[6,0,0],"C":[6,0,0],"new_shelf":1,"violation":0} | {"A":[0,0,3],"B":[6,0,0],"C":[6,0,0],"new_shelf":1,"violation":0} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0006|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 0, 3] B=[5, 0, 0] C=[6, 0, 0]
Arrivals=['S', 'P']
Arrival 1: S
Try A: can_place=0 state=[0, 0, 3]
Try B: can_place=1 state=[5, 0, 0]
Placed on B: new_state=[6, 0, 0]
Arrival 2: P
Try A: can_place=0 state=[0, 0, 3]
Try B: can_place=0 state=[6, 0, 0]
Try C: can_place=0 state=[6, 0, 0]
No place... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184839 | quantized_vllm | fuzz | fuzz_0007 | tool_decisions | 0.31 | 0 | null | null | 4 | 1 | 4 | 0 | 0 | 1 | 1 | 10 | ok | ["STOP","STOP","STOP","STOP"] | ["STOP","STOP","STOP","STOP"] | {"A":[0,1,0],"B":[4,2,0],"C":[5,1,0],"new_shelf":1,"violation":1} | {"A":[0,1,0],"B":[4,2,0],"C":[5,1,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0007|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 1, 0] B=[4, 2, 0] C=[5, 1, 0]
Arrivals=['H', 'P', 'S', 'P']
Arrival 1: H
Try A: can_place=0 state=[0, 1, 0]
Try B: can_place=0 state=[4, 2, 0]
Try C: can_place=0 state=[5, 1, 0]
No placement possible on A/B/C: placements=STOP
Arrival 2: P
Already stopped: placements=STOP
Arrival 3: S
Already s... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184840 | quantized_vllm | fuzz | fuzz_0008 | tool_decisions | 0.72 | 0 | null | null | 3 | 1 | 3 | 1 | 1 | 1 | 1 | 10 | ok | ["C","STOP","STOP"] | ["C","STOP","STOP"] | {"A":[0,0,5],"B":[2,4,0],"C":[5,1,0],"new_shelf":1,"violation":0} | {"A":[0,0,5],"B":[2,4,0],"C":[5,1,0],"new_shelf":1,"violation":0} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0008|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 0, 5] B=[2, 4, 0] C=[4, 1, 0]
Arrivals=['S', 'S', 'H']
Arrival 1: S
Try A: can_place=0 state=[0, 0, 5]
Try B: can_place=0 state=[2, 4, 0]
Try C: can_place=1 state=[4, 1, 0]
Placed on C: new_state=[5, 1, 0]
Arrival 2: S
Try A: can_place=0 state=[0, 0, 5]
Try B: can_place=0 state=[2, 4, 0]
Try... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184841 | quantized_vllm | fuzz | fuzz_0009 | tool_decisions | 0.76 | 0 | null | null | 4 | 1 | 4 | 1 | 1 | 1 | 1 | 10 | ok | ["C","STOP","STOP","STOP"] | ["C","STOP","STOP","STOP"] | {"A":[6,0,0],"B":[1,5,0],"C":[0,0,6],"new_shelf":1,"violation":1} | {"A":[6,0,0],"B":[1,5,0],"C":[0,0,6],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0009|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[6, 0, 0] B=[1, 5, 0] C=[0, 0, 5]
Arrivals=['H', 'S', 'P', 'P']
Arrival 1: H
Try A: can_place=0 state=[6, 0, 0]
Try B: can_place=0 state=[1, 5, 0]
Try C: can_place=1 state=[0, 0, 5]
Placed on C: new_state=[0, 0, 6]
Arrival 2: S
Try A: can_place=0 state=[6, 0, 0]
Try B: can_place=0 state=[1, 5, 0]
... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184841 | quantized_vllm | fuzz | fuzz_0010 | tool_decisions | 0.71 | 0 | null | null | 5 | 1 | 5 | 1 | 1 | 1 | 1 | 10 | ok | ["A","STOP","STOP","STOP","STOP"] | ["A","STOP","STOP","STOP","STOP"] | {"A":[3,3,0],"B":[6,0,0],"C":[0,5,0],"new_shelf":1,"violation":1} | {"A":[3,3,0],"B":[6,0,0],"C":[0,5,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0010|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[3, 2, 0] B=[6, 0, 0] C=[0, 5, 0]
Arrivals=['P', 'H', 'P', 'S', 'P']
Arrival 1: P
Try A: can_place=1 state=[3, 2, 0]
Placed on A: new_state=[3, 3, 0]
Arrival 2: H
Try A: can_place=0 state=[3, 3, 0]
Try B: can_place=0 state=[6, 0, 0]
Try C: can_place=0 state=[0, 5, 0]
No placement possible on A/B/C... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184842 | quantized_vllm | fuzz | fuzz_0011 | tool_decisions | 0.74 | 0 | null | null | 5 | 1 | 5 | 1 | 1 | 1 | 1 | 10 | ok | ["A","STOP","STOP","STOP","STOP"] | ["A","STOP","STOP","STOP","STOP"] | {"A":[4,0,0],"B":[5,1,0],"C":[6,0,0],"new_shelf":1,"violation":1} | {"A":[4,0,0],"B":[5,1,0],"C":[6,0,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0011|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[3, 0, 0] B=[5, 1, 0] C=[6, 0, 0]
Arrivals=['S', 'H', 'P', 'H', 'H']
Arrival 1: S
Try A: can_place=1 state=[3, 0, 0]
Placed on A: new_state=[4, 0, 0]
Arrival 2: H
Try A: can_place=0 state=[4, 0, 0]
Try B: can_place=0 state=[5, 1, 0]
Try C: can_place=0 state=[6, 0, 0]
No placement possible on A/B/C... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184844 | quantized_vllm | fuzz | fuzz_0012 | tool_decisions | 1.63 | 0 | null | null | 4 | 1 | 4 | -1 | -1 | 1 | 1 | 10 | ok | ["A","A","A","C"] | ["A","A","A","C"] | {"A":[3,3,0],"B":[4,2,0],"C":[4,2,0],"new_shelf":0,"violation":1} | {"A":[3,3,0],"B":[4,2,0],"C":[4,2,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0012|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[1, 2, 0] B=[4, 2, 0] C=[4, 1, 0]
Arrivals=['S', 'S', 'P', 'P']
Arrival 1: S
Try A: can_place=1 state=[1, 2, 0]
Placed on A: new_state=[2, 2, 0]
Arrival 2: S
Try A: can_place=1 state=[2, 2, 0]
Placed on A: new_state=[3, 2, 0]
Arrival 3: P
Try A: can_place=1 state=[3, 2, 0]
Placed on A: new_state=[... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184845 | quantized_vllm | fuzz | fuzz_0013 | tool_decisions | 1.13 | 0 | null | null | 5 | 1 | 5 | 2 | 2 | 1 | 1 | 10 | ok | ["A","B","STOP","STOP","STOP"] | ["A","B","STOP","STOP","STOP"] | {"A":[2,4,0],"B":[1,5,0],"C":[0,6,0],"new_shelf":1,"violation":1} | {"A":[2,4,0],"B":[1,5,0],"C":[0,6,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0013|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[1, 4, 0] B=[0, 5, 0] C=[0, 6, 0]
Arrivals=['S', 'S', 'H', 'P', 'H']
Arrival 1: S
Try A: can_place=1 state=[1, 4, 0]
Placed on A: new_state=[2, 4, 0]
Arrival 2: S
Try A: can_place=0 state=[2, 4, 0]
Try B: can_place=1 state=[0, 5, 0]
Placed on B: new_state=[1, 5, 0]
Arrival 3: H
Try A: can_place=0 ... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184847 | quantized_vllm | fuzz | fuzz_0014 | tool_decisions | 1.9 | 0 | null | null | 6 | 1 | 6 | 4 | 4 | 1 | 1 | 10 | ok | ["B","B","A","A","STOP","STOP"] | ["B","B","A","A","STOP","STOP"] | {"A":[0,0,6],"B":[3,3,0],"C":[6,0,0],"new_shelf":1,"violation":1} | {"A":[0,0,6],"B":[3,3,0],"C":[6,0,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0014|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 0, 4] B=[3, 1, 0] C=[6, 0, 0]
Arrivals=['P', 'P', 'H', 'H', 'P', 'P']
Arrival 1: P
Try A: can_place=0 state=[0, 0, 4]
Try B: can_place=1 state=[3, 1, 0]
Placed on B: new_state=[3, 2, 0]
Arrival 2: P
Try A: can_place=0 state=[0, 0, 4]
Try B: can_place=1 state=[3, 2, 0]
Placed on B: new_state=[3... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184848 | quantized_vllm | fuzz | fuzz_0015 | tool_decisions | 0.8 | 0 | null | null | 2 | 1 | 2 | -1 | -1 | 1 | 1 | 10 | ok | ["C","A"] | ["C","A"] | {"A":[4,0,0],"B":[0,2,0],"C":[0,0,5],"new_shelf":0,"violation":1} | {"A":[4,0,0],"B":[0,2,0],"C":[0,0,5],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0015|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[3, 0, 0] B=[0, 2, 0] C=[0, 0, 4]
Arrivals=['H', 'S']
Arrival 1: H
Try A: can_place=0 state=[3, 0, 0]
Try B: can_place=0 state=[0, 2, 0]
Try C: can_place=1 state=[0, 0, 4]
Placed on C: new_state=[0, 0, 5]
Arrival 2: S
Try A: can_place=1 state=[3, 0, 0]
Placed on A: new_state=[4, 0, 0]
Final A=[4, ... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184848 | quantized_vllm | fuzz | fuzz_0016 | tool_decisions | 0.4 | 0 | null | null | 1 | 1 | 1 | -1 | -1 | 1 | 1 | 10 | ok | ["A"] | ["A"] | {"A":[0,0,6],"B":[1,0,0],"C":[6,0,0],"new_shelf":0,"violation":1} | {"A":[0,0,6],"B":[1,0,0],"C":[6,0,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0016|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 0, 5] B=[1, 0, 0] C=[6, 0, 0]
Arrivals=['H']
Arrival 1: H
Try A: can_place=1 state=[0, 0, 5]
Placed on A: new_state=[0, 0, 6]
Final A=[0, 0, 6] B=[1, 0, 0] C=[6, 0, 0] new_shelf=0 violation=1 | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184848 | quantized_vllm | fuzz | fuzz_0017 | tool_decisions | 0.31 | 0 | null | null | 4 | 1 | 4 | 0 | 0 | 1 | 1 | 10 | ok | ["STOP","STOP","STOP","STOP"] | ["STOP","STOP","STOP","STOP"] | {"A":[5,1,0],"B":[1,5,0],"C":[4,2,0],"new_shelf":1,"violation":1} | {"A":[5,1,0],"B":[1,5,0],"C":[4,2,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0017|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[5, 1, 0] B=[1, 5, 0] C=[4, 2, 0]
Arrivals=['S', 'P', 'P', 'S']
Arrival 1: S
Try A: can_place=0 state=[5, 1, 0]
Try B: can_place=0 state=[1, 5, 0]
Try C: can_place=0 state=[4, 2, 0]
No placement possible on A/B/C: placements=STOP
Arrival 2: P
Already stopped: placements=STOP
Arrival 3: P
Already s... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184849 | quantized_vllm | fuzz | fuzz_0018 | tool_decisions | 0.7 | 0 | null | null | 6 | 1 | 6 | 1 | 1 | 1 | 1 | 10 | ok | ["A","STOP","STOP","STOP","STOP","STOP"] | ["A","STOP","STOP","STOP","STOP","STOP"] | {"A":[0,0,6],"B":[3,3,0],"C":[5,0,0],"new_shelf":1,"violation":1} | {"A":[0,0,6],"B":[3,3,0],"C":[5,0,0],"new_shelf":1,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0018|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[0, 0, 5] B=[3, 3, 0] C=[5, 0, 0]
Arrivals=['H', 'H', 'S', 'S', 'P', 'H']
Arrival 1: H
Try A: can_place=1 state=[0, 0, 5]
Placed on A: new_state=[0, 0, 6]
Arrival 2: H
Try A: can_place=0 state=[0, 0, 6]
Try B: can_place=0 state=[3, 3, 0]
Try C: can_place=0 state=[5, 0, 0]
No placement possible on ... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184850 | quantized_vllm | fuzz | fuzz_0019 | tool_decisions | 1.19 | 0 | null | null | 3 | 1 | 3 | -1 | -1 | 1 | 1 | 10 | ok | ["B","C","C"] | ["B","C","C"] | {"A":[6,0,0],"B":[1,5,0],"C":[5,1,0],"new_shelf":0,"violation":1} | {"A":[6,0,0],"B":[1,5,0],"C":[5,1,0],"new_shelf":0,"violation":1} | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|fuzz|fuzz_0019|tool_decisions | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | Initial A=[6, 0, 0] B=[1, 4, 0] C=[3, 1, 0]
Arrivals=['P', 'S', 'S']
Arrival 1: P
Try A: can_place=0 state=[6, 0, 0]
Try B: can_place=1 state=[1, 4, 0]
Placed on B: new_state=[1, 5, 0]
Arrival 2: S
Try A: can_place=0 state=[6, 0, 0]
Try B: can_place=0 state=[1, 5, 0]
Try C: can_place=1 state=[3, 1, 0]
Pla... | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184850 | quantized_vllm | mcq | mcq_01 | null | 0.02 | 0 | B | null | null | 1 | null | null | null | 1 | null | 1 | ok | B | B | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|mcq|mcq_01| | null | null | B | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184850 | quantized_vllm | mcq | mcq_02 | null | 0.02 | 0 | A | null | null | 0 | null | null | null | 0 | null | 0 | wrong_choice got=A | B | A | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|mcq|mcq_02| | null | null | A | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184850 | quantized_vllm | mcq | mcq_03 | null | 0.02 | 0 | C | null | null | 1 | null | null | null | 1 | null | 1 | ok | C | C | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|mcq|mcq_03| | null | null | C | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184850 | quantized_vllm | mcq | mcq_04 | null | 0.02 | 0 | B | null | null | 1 | null | null | null | 1 | null | 1 | ok | B | B | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|mcq|mcq_04| | null | null | B | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184850 | quantized_vllm | mcq | mcq_05 | null | 0.02 | 0 | A | null | null | 0 | null | null | null | 0 | null | 0 | wrong_choice got=A | B | A | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|mcq|mcq_05| | null | null | A | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184851 | quantized_vllm | toolcall | tool_01 | null | 0.48 | 0 | {
"tool_name": "add",
"args": {
"a": 2,
"b": 3
}
}
5 | null | null | null | null | null | null | 1 | null | 11 | ok | 5 | 5 | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|toolcall|tool_01| | null | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | 1 | 1 | null | ok | 1 | null | {"tool_name": "add", "args": {"a": 2, "b": 3}} | {
"tool_name": "add",
"args": {
"a": 2,
"b": 3
}
} | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184851 | quantized_vllm | toolcall | tool_02 | null | 0.49 | 0 | {
"tool_name": "add",
"args": {
"a": 10,
"b": -4
}
}
6 | null | null | null | null | null | null | 1 | null | 11 | ok | 6 | 6 | null | null | null | null | null | null | null | null | null | null | null | null | null | quantized_vllm|toolcall|tool_02| | null | null | null | null | ok | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | null | 1 | 1 | null | ok | 1 | null | {"tool_name": "add", "args": {"a": 10, "b": -4}} | {
"tool_name": "add",
"args": {
"a": 10,
"b": -4
}
} | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184852 | quantized_vllm | toolcall_only | toolonly_01 | null | 0.3 | 0 | {"action": "add", "args": {"x": 5, "y": 10}} | null | null | null | null | null | null | null | null | 2 | schema_error | null | null | null | null | null | 1 | null | null | null | null | null | null | null | null | null | quantized_vllm|toolcall_only|toolonly_01| | null | null | null | null | null | null | {"a":5,"b":10} | null | null | null | null | null | null | {"a":5,"b":10} | null | null | null | null | null | null | null | 1 | null | Additional properties are not allowed ('action' was unexpected); {'action': 'add', 'args': {'x': 5, 'y': 10}} is not valid under any of the given schemas; {'x': 5, 'y': 10} is not valid under any of the given schemas | 0 | null | null | null | null | null | 1 | null | null | null | null | null | null |
Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555 | 20260629_184852 | quantized_vllm | toolcall_only | toolonly_02 | null | 0.32 | 0 | {"action": "add", "args": {"x": 25, "y": 75}} | null | null | null | null | null | null | null | null | 2 | schema_error | null | null | null | null | null | 1 | null | null | null | null | null | null | null | null | null | quantized_vllm|toolcall_only|toolonly_02| | null | null | null | null | null | null | {"a":25,"b":75} | null | null | null | null | null | null | {"a":25,"b":75} | null | null | null | null | null | null | null | 1 | null | Additional properties are not allowed ('action' was unexpected); {'action': 'add', 'args': {'x': 25, 'y': 75}} is not valid under any of the given schemas; {'x': 25, 'y': 75} is not valid under any of the given schemas | 0 | null | null | null | null | null | 1 | null | null | null | null | null | null |
quant_eval v7.21 — Per-Case Evaluation Results and Run Provenance
Supplementary evidence for the whitepaper quant_eval: A Behavioral Evaluation Harness for Full-Weight and Quantized Large Language Models.
Author: Patrick Hill, PBH Applied Systems, LLC
ORCID: 0009-0008-3662-1681
Licence: CC BY 4.0
Concept DOI (all versions): 10.5281/zenodo.22851375
Version DOI (this deposit): 10.5281/zenodo.22851376
What this deposit is
Every quantitative result reported in the whitepaper comes from two evaluation runs of quant_eval version 7.21. This deposit contains the complete per-case output of both runs and the provenance manifest each run wrote, so that any pass rate, bucket score, or latency figure in the paper can be recomputed from source and any individual failure inspected at the row level.
It contains 126 scored cases across three inference backends, 72 recorded signals per case.
What this deposit is not
This is not part of the quant_eval public corpus. The corpus (datasets D0–D7) reports production evaluations under quant_eval v7.22 and is deposited separately. Nothing here should be merged with it, compared against it, or read as superseding it.
The methodology used here was retired. Both runs used the fast_gate screening profile: a
small per-family case count intended to triage models quickly, not to establish statistical
confidence. The harness itself marks every family verdict in these runs provisional (n<10),
and that qualifier is not decoration. fast_gate was retired immediately after the whitepaper
was written and is not the methodology behind any published corpus dataset.
These are screening results, not production results. They are published because the whitepaper rests on them and a reader is entitled to check the arithmetic — not because they are a fit basis for a deployment decision about either model.
No model weights are redistributed here. The evaluated models remain under their own licences.
Files
| File | Bytes | Contents |
|---|---|---|
comparison_results_v7_21_Mistral_Nemo_Instruct_2407_20260211_022944.csv |
68,719 | 84 rows — the side-by-side run, one row per case per runner |
comparison_results_v7_21_Mistral_Nemo_Instruct_2407_AWQ_20260529_152601_20260629_184555.csv |
36,858 | 42 rows — the AWQ run |
run_manifest_20260211_022944.json |
3,565 | provenance for the side-by-side run (sanitized; see below) |
run_manifest_20260629_184555.json |
2,451 | provenance for the AWQ run (sanitized; see below) |
README.md |
— | this file |
LICENSE |
— | CC BY 4.0 grant and suggested attribution |
CITATION.cff |
— | machine-readable citation metadata |
Both manifests were written by the harness under the filename run_manifest.json. They are
renamed here by run identifier because a flat archive cannot hold two files of the same name;
the original filename is recorded inside each file.
The two runs
| Side-by-side run | AWQ run | |
|---|---|---|
| Run ID | 20260211_022944 |
20260629_184555 |
| Date | 2026-02-11 | 2026-06-29 |
| Model | mistralai/Mistral-Nemo-Instruct-2407 |
local AWQ build Mistral-Nemo-Instruct-2407-AWQ_20260529_152601 |
| Runners | full_weight_transformers (FP16, Hugging Face Transformers)quantized_llama_cpp (Q4_K_M, llama.cpp) |
quantized_vllm (W4A16 AWQ, vLLM) |
| Rows | 84 | 42 |
| Harness version | 7.21 | 7.21 |
| Seed | 42 | 42 |
Both runs executed the identical fixture set:
golden_oracle_fixtures_v7_21__sha256_6d71a0b9147c.json
sha256 = 6d71a0b9147c079371b02a94f3c149eb78a6adc03dc16ff6833b964fbf4174f0
Each runner saw the same 42 cases across eight families: json (4), json_multistep (5),
stateful_followup (2), mixed_brief_json (2), mcq (5), toolcall (2), toolcall_only (2),
fuzz (20). The authored battery is 15 cases; mcq, toolcall and the 20-case fuzz
property-based regression suite bring each runner's total to 42.
Decoding
Both manifests record "seed": 42. The decoding temperature is not recorded in these
files. The harness applied sampled decoding at the temperature recommended on the
Mistral-Nemo-Instruct-2407 model card, 0.3, together with that card's other recommended
parameters.
The CSVs corroborate that sampling — not greedy decoding — was in use. All 15 json_multistep
rows across both runs carry mode = best_of_k, best_of_k_k = 3, best_of_k_used = 1, with
rotating per-case seeds (1337, 1338, 1437, 1537, 1637, 1638, 1737). In two of those rows the
selector chose candidate index 1 rather than 0 — ms_easy_01 under
full_weight_transformers (recorded similarity 0.682) and ms_med_02 under
quantized_llama_cpp (0.000). Greedy decoding yields one deterministic completion per prompt,
so three distinct candidates could not have existed to choose among.
Results
Pass rate is the fraction of rows in a family where every gating signal equals 1:
| Family | Gating signals |
|---|---|
json_multistep |
schema_ok, checks_consistent_ok, stop_semantics_ok, oracle_equiv_ok |
stateful_followup |
turn1_parse_ok, turn2_parse_ok, turn1_exact_match, turn2_exact_match |
toolcall_only |
tool_name_ok, args_ok |
mixed_brief_json |
answer_line_ok, json_parse_ok, schema_ok |
toolcall |
stage1_tool_parse_ok, stage1_tool_schema_ok |
json, fuzz and mcq report bucket_score rather than a conjunctive pass rate; their pass
column reads n/a.
Side-by-side run 20260211_022944
| Family | N | FP16 pass | FP16 bucket | Q4_K_M pass | Q4_K_M bucket |
|---|---|---|---|---|---|
| json_multistep | 5 | 0.600 | 1.800 | 0.400 | 1.600 |
| stateful_followup | 2 | 1.000 | 2.000 | 1.000 | 2.000 |
| toolcall_only | 2 | 1.000 | 2.000 | 0.000 | 1.000 |
| mixed_brief_json | 2 | 1.000 | 2.000 | 1.000 | 2.000 |
| json | 4 | n/a | 10.000 | n/a | 10.000 |
| fuzz | 20 | n/a | 10.000 | n/a | 10.000 |
| toolcall | 2 | 1.000 | 11.000 | 1.000 | 5.500 |
| mcq | 5 | n/a | 0.600 | n/a | 0.400 |
AWQ run 20260629_184555
| Family | N | W4A16 pass | W4A16 bucket |
|---|---|---|---|
| json_multistep | 5 | 0.400 | 1.800 |
| stateful_followup | 2 | 1.000 | 2.000 |
| toolcall_only | 2 | 1.000 | 2.000 |
| mixed_brief_json | 2 | 1.000 | 2.000 |
| json | 4 | n/a | 10.000 |
| fuzz | 20 | n/a | 10.000 |
| toolcall | 2 | 1.000 | 11.000 |
| mcq | 5 | n/a | 0.600 |
Every figure above was recomputed directly from the deposited CSVs and matches the rollup the harness produced at run time.
Wall-clock timings, and why they are not a precision comparison
Mean per-case latency over all 42 cases, computed from the secs column:
| Runner | Precision | Backend | Mean s/case |
|---|---|---|---|
full_weight_transformers |
FP16 | Hugging Face Transformers | 30.244 |
quantized_llama_cpp |
Q4_K_M | llama.cpp | 1.417 |
quantized_vllm |
W4A16 AWQ | vLLM | 0.936 |
These three numbers span three different inference backends. The spread between them combines a backend change with a precision change and cannot be attributed to quantization alone. They are evaluation-harness measurements under a fixed configuration, not throughput benchmarks, and they are published here so the whitepaper's figures can be traced to their source — not as a performance claim.
Recorded model artifacts
The side-by-side manifest records both GGUF artifacts with byte counts and SHA-256 digests:
| Kind | Quant type | File | Bytes | SHA-256 |
|---|---|---|---|---|
gguf_f16 |
— | hf_snapshot_F16.gguf |
24,504,279,808 | cc7b8c5c3f129ad32aee562017b5e56f1284ee7d92445be292c644a42b3c9556 |
gguf_quant |
Q4_K_M | hf_snapshot_Q4_K_M.gguf |
7,477,207,808 | 5765024ff3361f6dc5b590b963b378bd2e87ac95eabe5823a08a3ad336b498c9 |
That is a 3.277× reduction in stored artifact size — a property of the files themselves, independent of backend.
The AWQ manifest records no artifacts ("artifacts": []). No file digest exists for the
AWQ model. Artifact-level verification is therefore possible for the FP16/Q4_K_M pair and not
for the AWQ lane.
Cache reuse in the side-by-side run
The full-weight lane of run 20260211_022944 did not re-execute. Its results were reused from
the cache written by an earlier run. The manifest records this:
"full_weight_transformers": {
"cache_reuse_path": "<WORK_ROOT>/runs/.../Mistral_Nemo_Instruct_2407_20260211_014604/full_weight_cache.json",
"cache_write_enabled": 0,
"reused_cache": 1
}
The CSV corroborates it. The 42 full-weight rows carry only two distinct timestamp values,
20260211_023003 and 20260211_023004 — one second apart. Forty-two cases averaging 30.24 s
cannot execute in two seconds; they were written out from cache in a single pass. The 42
quantized rows span 24 distinct timestamps across roughly ninety seconds of real execution.
The full-weight rows carry run_id = 20260211_022944 while the quantized rows carry
run_id = Mistral_Nemo_Instruct_2407_20260211_022944. The two spellings are an artifact of the
two lanes being written by different code paths and are left exactly as recorded.
The comparison remains sound — the fixture set, the seed and the prompts are identical across
the source run and the side-by-side run — but the timings and scores in the full-weight lane
were measured in run 20260211_014604, and a reader should know that.
Manifest sanitization
The deposited manifests are not byte-identical to the files the harness wrote. quant_eval is
proprietary and unpublished; its directory layout is not disclosed. Each machine-local
filesystem root was replaced with a symbolic token — <WORK_ROOT>, <CATALOG_ROOT>,
<TOOLCHAIN_ROOT>, <MODEL_LIBRARY>, <ALLOWED_ROOT> — and the path structure below each
root was left verbatim, so cross-run references such as the cache source above remain
traceable.
Fourteen strings were altered in the side-by-side manifest and ten in the AWQ manifest. Each
file carries a _sanitization block recording the tool, the SHA-256 of the unmodified source
file, and the count of altered strings.
No run identifier, timestamp, seed, fixture hash, artifact byte count, artifact digest, toolchain digest, runner role, quantization type, or cache-reuse field was altered. The transformation was applied programmatically under two abort conditions: any surviving machine-local path, or any change to a field on the preserved list.
The CSVs are deposited exactly as the harness wrote them. Nothing in them was modified, reordered, or removed.
CSV schema
Both files share an identical 72-column header. Orientation for the columns that carry most of the weight:
| Column | Meaning |
|---|---|
run_id, timestamp |
run and per-case execution identifiers |
runner |
which backend produced the row |
family, case_id |
battery family and case within it |
mode |
scoring mode (best_of_k, 2turn, tool_decisions, …) |
secs |
wall-clock seconds for the case |
output_raw |
the model's verbatim output |
bucket_score |
family-specific auxiliary score, scale documented per family |
detail |
failure reason, or ok |
expected_* / got_* |
oracle value and observed value, paired |
*_ok |
the individual scored signals |
best_of_k_* |
candidate-selection record |
bucket_score scales are family-specific and are not comparable across families. For
json_multistep a 10 requires plan_exact_match, checks_ok and final_match_reported
together; otherwise the score is the number of leading plan steps matching the oracle. mcq
runs 0–1. The rest run 0–10.
For json_multistep, Tier 1 (gating) covers planning feasibility and externally verified
outcome correctness; Tier 2 (final_consistent_ok, final_match_reported) tracks final-state
reporting discipline and does not gate a Tier 1 pass. Both Tier 2 rates are 0.000 across all
three runners in these runs.
How to verify
import csv, statistics
csv.field_size_limit(2**31 - 1)
rows = list(csv.DictReader(
open("comparison_results_v7_21_Mistral_Nemo_Instruct_2407_20260211_022944.csv",
newline="", encoding="utf-8")))
gating = ["schema_ok", "checks_consistent_ok", "stop_semantics_ok", "oracle_equiv_ok"]
for runner in ("full_weight_transformers", "quantized_llama_cpp"):
sel = [r for r in rows if r["runner"] == runner and r["family"] == "json_multistep"]
passed = sum(1 for r in sel if all(r[c] == "1" for c in gating))
print(runner, passed / len(sel)) # 0.6 then 0.4
for runner in ("full_weight_transformers", "quantized_llama_cpp"):
sel = [r for r in rows if r["runner"] == runner]
print(runner, round(statistics.mean(float(r["secs"]) for r in sel), 3)) # 30.244 / 1.417
Model artifacts can be checked against the digests in run_manifest_20260211_022944.json,
where the source files are available.
Limitations
- Sample sizes are small by design. Four of eight families have N ≤ 5 per runner. The harness labels every verdict in these runs provisional; treat them as screening signal.
- Backends differ across the three runners. No same-backend precision comparison exists in this data.
- One worked model. Everything here concerns Mistral-Nemo-Instruct-2407. Nothing generalizes to other model families without re-running.
- The full-weight lane was cache-reused, as recorded above.
- No artifact digest exists for the AWQ model.
- Fixture contents are not included. The CSVs record each case's expected value and the model's output, not the prompts. The fixture file is identified by SHA-256 only.
- The scoring code is not published. quant_eval is proprietary. These files support recomputation of the reported statistics from recorded signals; they do not permit re-deriving the signals from raw outputs.
Citation
Hill, Patrick. quant_eval v7.21 — Per-Case Evaluation Results and Run Provenance. PBH Applied Systems, LLC, 2026. CC BY 4.0. DOI:
10.5281/zenodo.22851375
See CITATION.cff for machine-readable metadata.
Related
- Whitepaper — quant_eval: A Behavioral Evaluation Harness for Full-Weight and Quantized
Large Language Models. This deposit is its evidence base. Concept DOI
10.5281/zenodo.22851592. - quant_eval public corpus (D0–D7) — production evaluations under v7.22, deposited
separately. Concept DOI
10.5281/zenodo.22009419. Different harness version, different methodology, different corpus. Do not conflate the two.
PBH Applied Systems, LLC — Oklahoma City, Oklahoma.
- Downloads last month
- 35