{ "model": "transformer", "seed": 0, "params": null, "val_selected": { "ckpt": "best.pt", "params": 407947, "val_token_acc": 0.9545454545454546, "val_exact_match": 0.8666666666666667, "note": "val-selected checkpoint (early stop / best-on-val per prereg \u2014 val EM here is selection-holed, not an untouched test estimate" }, "final": { "ckpt": "last.pt", "params": 407947, "val_token_acc": 0.8863636363636364, "val_exact_match": 0.7, "note": "final checkpoint, no val selection", "probe": { "code": "P4", "acc": 0.0, "correct": 0, "total": 100, "errors": [ { "n": 101, "target": 103, "pred": 101 }, { "n": 102, "target": 103, "pred": 101 }, { "n": 103, "target": 107, "pred": 101 }, { "n": 104, "target": 107, "pred": 101 }, { "n": 105, "target": 107, "pred": 171 }, { "n": 106, "target": 107, "pred": 101 }, { "n": 107, "target": 109, "pred": 101 }, { "n": 108, "target": 109, "pred": 101 }, { "n": 109, "target": 113, "pred": 101 }, { "n": 110, "target": 113, "pred": 101 }, { "n": 111, "target": 113, "pred": 101 }, { "n": 112, "target": 113, "pred": 101 }, { "n": 113, "target": 127, "pred": 101 }, { "n": 114, "target": 127, "pred": 171 }, { "n": 115, "target": 127, "pred": 171 }, { "n": 116, "target": 127, "pred": 171 }, { "n": 117, "target": 127, "pred": 101 }, { "n": 118, "target": 127, "pred": 101 }, { "n": 119, "target": 127, "pred": 20 }, { "n": 120, "target": 127, "pred": 20 }, { "n": 121, "target": 127, "pred": 101 }, { "n": 122, "target": 127, "pred": 20 }, { "n": 123, "target": 127, "pred": 20 }, { "n": 124, "target": 127, "pred": 101 }, { "n": 125, "target": 127, "pred": 101 }, { "n": 126, "target": 127, "pred": 20 }, { "n": 127, "target": 131, "pred": 20 }, { "n": 128, "target": 131, "pred": 20 }, { "n": 129, "target": 131, "pred": 20 }, { "n": 130, "target": 131, "pred": 90 }, { "n": 131, "target": 137, "pred": 97 }, { "n": 132, "target": 137, "pred": 37 }, { "n": 133, "target": 137, "pred": 97 }, { "n": 134, "target": 137, "pred": 97 }, { "n": 135, "target": 137, "pred": 37 }, { "n": 136, "target": 137, "pred": 97 }, { "n": 137, "target": 139, "pred": 101 }, { "n": 138, "target": 139, "pred": 101 }, { "n": 139, "target": 149, "pred": 101 }, { "n": 140, "target": 149, "pred": 101 }, { "n": 141, "target": 149, "pred": 101 }, { "n": 142, "target": 149, "pred": 101 }, { "n": 143, "target": 149, "pred": 171 }, { "n": 144, "target": 149, "pred": 171 }, { "n": 145, "target": 149, "pred": 171 }, { "n": 146, "target": 149, "pred": 171 }, { "n": 147, "target": 149, "pred": 50 }, { "n": 148, "target": 149, "pred": 50 }, { "n": 149, "target": 151, "pred": 60 }, { "n": 150, "target": 151, "pred": 50 }, { "n": 151, "target": 157, "pred": 50 }, { "n": 152, "target": 157, "pred": 50 }, { "n": 153, "target": 157, "pred": 50 }, { "n": 154, "target": 157, "pred": 50 }, { "n": 155, "target": 157, "pred": 171 }, { "n": 156, "target": 157, "pred": 50 }, { "n": 157, "target": 163, "pred": 50 }, { "n": 158, "target": 163, "pred": 50 }, { "n": 159, "target": 163, "pred": 60 }, { "n": 160, "target": 163, "pred": 101 }, { "n": 161, "target": 163, "pred": 101 }, { "n": 162, "target": 163, "pred": 171 }, { "n": 163, "target": 167, "pred": 97 }, { "n": 164, "target": 167, "pred": 97 }, { "n": 165, "target": 167, "pred": 97 }, { "n": 166, "target": 167, "pred": 97 }, { "n": 167, "target": 173, "pred": 101 }, { "n": 168, "target": 173, "pred": 101 }, { "n": 169, "target": 173, "pred": 50 }, { "n": 170, "target": 173, "pred": 70 }, { "n": 171, "target": 173, "pred": 70 }, { "n": 172, "target": 173, "pred": 70 }, { "n": 173, "target": 179, "pred": 79 }, { "n": 174, "target": 179, "pred": 77 }, { "n": 175, "target": 179, "pred": 77 }, { "n": 176, "target": 179, "pred": 70 }, { "n": 177, "target": 179, "pred": 101 }, { "n": 178, "target": 179, "pred": 70 }, { "n": 179, "target": 181, "pred": 80 }, { "n": 180, "target": 181, "pred": 80 }, { "n": 181, "target": 191, "pred": 80 }, { "n": 182, "target": 191, "pred": 80 }, { "n": 183, "target": 191, "pred": 80 }, { "n": 184, "target": 191, "pred": 87 }, { "n": 185, "target": 191, "pred": 87 }, { "n": 186, "target": 191, "pred": 87 }, { "n": 187, "target": 191, "pred": 101 }, { "n": 188, "target": 191, "pred": 101 }, { "n": 189, "target": 191, "pred": 87 }, { "n": 190, "target": 191, "pred": 97 }, { "n": 191, "target": 193, "pred": 97 }, { "n": 192, "target": 193, "pred": 97 }, { "n": 193, "target": 197, "pred": 97 }, { "n": 194, "target": 197, "pred": 97 }, { "n": 195, "target": 197, "pred": 97 }, { "n": 196, "target": 197, "pred": 97 }, { "n": 197, "target": 199, "pred": 101 }, { "n": 198, "target": 199, "pred": 101 }, { "n": 199, "target": 211, "pred": 101 }, { "n": 200, "target": 211, "pred": 101 } ], "flagged_errors": [], "flagged_pred_fraction": 0.0, "distinct_flagged_preds": 0, "easy_total": 60, "easy_wrong": 60, "easy_err_rate": 1.0 }, "per_example": [ { "n": "3", "target": 5, "pred": 3, "ok": false }, { "n": "4", "target": 5, "pred": 7, "ok": false }, { "n": "8", "target": 11, "pred": 1, "ok": false }, { "n": "9", "target": 11, "pred": 31, "ok": false }, { "n": "10", "target": 11, "pred": 13, "ok": false }, { "n": "12", "target": 13, "pred": 13, "ok": true }, { "n": "13", "target": 17, "pred": 17, "ok": true }, { "n": "15", "target": 17, "pred": 17, "ok": true }, { "n": "18", "target": 19, "pred": 19, "ok": true }, { "n": "21", "target": 23, "pred": 23, "ok": true }, { "n": "24", "target": 29, "pred": 29, "ok": true }, { "n": "25", "target": 29, "pred": 29, "ok": true }, { "n": "26", "target": 29, "pred": 29, "ok": true }, { "n": "31", "target": 37, "pred": 31, "ok": false }, { "n": "32", "target": 37, "pred": 37, "ok": true }, { "n": "39", "target": 41, "pred": 31, "ok": false }, { "n": "43", "target": 47, "pred": 47, "ok": true }, { "n": "49", "target": 53, "pred": 61, "ok": false }, { "n": "50", "target": 53, "pred": 53, "ok": true }, { "n": "54", "target": 59, "pred": 59, "ok": true }, { "n": "56", "target": 59, "pred": 59, "ok": true }, { "n": "61", "target": 67, "pred": 61, "ok": false }, { "n": "65", "target": 67, "pred": 67, "ok": true }, { "n": "68", "target": 71, "pred": 71, "ok": true }, { "n": "78", "target": 79, "pred": 79, "ok": true }, { "n": "82", "target": 83, "pred": 83, "ok": true }, { "n": "84", "target": 89, "pred": 89, "ok": true }, { "n": "88", "target": 89, "pred": 89, "ok": true }, { "n": "92", "target": 97, "pred": 97, "ok": true }, { "n": "95", "target": 97, "pred": 97, "ok": true } ] }, "signature": { "code": "O-PARTIAL", "sat_start_eval": 0, "val_hi_eval_idx": null, "transition_width_evals": null, "n_evals": 1000, "final_train_em": 1.0, "final_val_em": 0.7, "max_train_em": 1.0 } }