File size: 2,263 Bytes
a06d854 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 | {
"benchmark": "deepresearch9k-frontier-v1",
"blocked_docids_sha256": "a4198790802f951af4c183db299cb526dd1d6592529f1f1890833d0b7108343d",
"by_calibration_tier": {
"challenge": 128,
"frontier": 128
},
"by_calibration_tier_and_selection_bucket": {
"challenge/hard_open": 28,
"challenge/reformulate": 100,
"frontier/hard_open": 28,
"frontier/reformulate": 100
},
"by_cell": {
"hard_open/D1/teacher_false": 12,
"hard_open/D1/teacher_true": 25,
"hard_open/D2/teacher_true": 1,
"hard_open/D3/teacher_false": 12,
"hard_open/D3/teacher_true": 6,
"reformulate/D1/teacher_false": 8,
"reformulate/D1/teacher_true": 54,
"reformulate/D2/teacher_false": 1,
"reformulate/D2/teacher_true": 5,
"reformulate/D3/teacher_false": 58,
"reformulate/D3/teacher_true": 74
},
"by_selection_bucket": {
"hard_open": 56,
"reformulate": 200
},
"challenge_eligible_by_selection_bucket": {
"hard_open": 37,
"reformulate": 224
},
"eligible_by_selection_bucket": {
"hard_open": 29,
"reformulate": 102
},
"max_fast_successes": 0,
"min_mean_turns": 3.0,
"pilot_audit": "selection/pilot_audit.jsonl",
"pilot_audit_sha256": "6dba18a6eee7fd5b2e068c28ff5acbd3d44581ffee48d14047944458f6101502",
"pilot_data": "selection/pilot_data.jsonl",
"pilot_data_sha256": "0359ca036dcb24a4d8397bae1755b000bc1af342e58fe84f93efe158a838dfbb",
"pilot_replicas": 3,
"pilot_results": "selection/pilot_results.jsonl",
"pilot_results_sha256": "eaa26af31d5673e93032a84031286b2dd9643c057914dd21af574abcb387c031",
"questions": 256,
"questions_sha256": "6c1bbdfad639b84224976d4a49d28e1eb060cb0a38f3f7284b9050b6963e7a85",
"rejected": {
"blank_answer": 111,
"non_finish": 3,
"one_step_success": 23,
"too_easy": 101,
"too_few_turns": 9,
"too_hard": 294
},
"seed": 20260810,
"selection_note": "Internal model-calibrated frontier set. Selection used repeated pre-RL base outputs; final scores must use fresh rollouts and are not a clean held-out benchmark estimate.",
"source_candidates": "selection/candidates.jsonl",
"source_candidates_sha256": "a893a0b7e79174e423a2a40deb1d3857df01313041e377e91c7f4920a73790c5",
"success_range": [
1,
2
]
}
|