1{2 "system": "p4v2_gr_s2",3 "split": "test_compose",4 "checkpoint": "experiments/runs/p4v2_gr_seed02_20261030001749/checkpoint",5 "checkpoint_sha256": "13e35362d3e2e4c169299b94379e544e3b64dde1d8cb8d7b0c5c32346d279037",6 "config_sha256": "0d380d727de6614b17db8524bf8e936a06211456dbd6fc7d3324b5d06e75fe70",7 "code_commit": "unknown",8 "device": "mps",9 "precision": "fp32",10 "evaluated_at": "2026-10-30T04:06:23-0400",11 "data_sha256": "06620e1cf60b7250aac71f4fb5ee74e3705f4b9aab102e22b5c5bdbed76c4ae2",12 "exact_match": {13 "budget0": {14 "accuracy": 0.046296296296296294,15 "n": 540,16 "by_family": {17 "compose": 0.04629629629629629418 },19 "edits_per_token": 0.020 },21 "forced_b2": {22 "accuracy": 0.046296296296296294,23 "n": 540,24 "by_family": {25 "compose": 0.04629629629629629426 },27 "edits_per_token": 2.028 }29 },30 "seconds": 1.931}