Spaces:
Running
Running
| { | |
| "model_type": "mamba_ssm", | |
| "tau_90": 0.9, | |
| "tau_99": 0.99, | |
| "N_prompts": 25, | |
| "dist_short": 10, | |
| "dist_long": 1000, | |
| "vocab_low": 1000, | |
| "vocab_high": 49000, | |
| "n_prefix": 20, | |
| "ref_logp_d10": { | |
| "mean": -6.946, | |
| "std": 3.4195 | |
| }, | |
| "baseline_logp_d1000": { | |
| "mean": -11.8226, | |
| "std": 2.9865 | |
| }, | |
| "logp_gap": 4.8765, | |
| "layer_sweep": { | |
| "0": { | |
| "recovery_pct": 0.0, | |
| "logp_transplant_mean": -11.8226, | |
| "logp_transplant_std": 2.9865 | |
| }, | |
| "1": { | |
| "recovery_pct": -0.1, | |
| "logp_transplant_mean": -11.8275, | |
| "logp_transplant_std": 2.9928 | |
| }, | |
| "2": { | |
| "recovery_pct": 0.37, | |
| "logp_transplant_mean": -11.8047, | |
| "logp_transplant_std": 3.1193 | |
| }, | |
| "3": { | |
| "recovery_pct": 0.48, | |
| "logp_transplant_mean": -11.799, | |
| "logp_transplant_std": 3.0748 | |
| }, | |
| "4": { | |
| "recovery_pct": 0.64, | |
| "logp_transplant_mean": -11.7912, | |
| "logp_transplant_std": 3.0539 | |
| }, | |
| "5": { | |
| "recovery_pct": 0.26, | |
| "logp_transplant_mean": -11.8099, | |
| "logp_transplant_std": 3.0662 | |
| }, | |
| "6": { | |
| "recovery_pct": 0.77, | |
| "logp_transplant_mean": -11.7848, | |
| "logp_transplant_std": 3.1346 | |
| }, | |
| "7": { | |
| "recovery_pct": 0.79, | |
| "logp_transplant_mean": -11.7839, | |
| "logp_transplant_std": 3.1639 | |
| }, | |
| "8": { | |
| "recovery_pct": 0.55, | |
| "logp_transplant_mean": -11.7959, | |
| "logp_transplant_std": 3.1798 | |
| }, | |
| "9": { | |
| "recovery_pct": 0.14, | |
| "logp_transplant_mean": -11.8155, | |
| "logp_transplant_std": 3.1905 | |
| }, | |
| "10": { | |
| "recovery_pct": 0.03, | |
| "logp_transplant_mean": -11.8213, | |
| "logp_transplant_std": 3.1787 | |
| }, | |
| "11": { | |
| "recovery_pct": 0.28, | |
| "logp_transplant_mean": -11.8089, | |
| "logp_transplant_std": 3.2391 | |
| }, | |
| "12": { | |
| "recovery_pct": 0.1, | |
| "logp_transplant_mean": -11.8177, | |
| "logp_transplant_std": 3.1534 | |
| }, | |
| "13": { | |
| "recovery_pct": -0.14, | |
| "logp_transplant_mean": -11.8294, | |
| "logp_transplant_std": 3.1626 | |
| }, | |
| "14": { | |
| "recovery_pct": -0.2, | |
| "logp_transplant_mean": -11.8322, | |
| "logp_transplant_std": 3.1637 | |
| }, | |
| "15": { | |
| "recovery_pct": 0.04, | |
| "logp_transplant_mean": -11.8206, | |
| "logp_transplant_std": 3.2057 | |
| }, | |
| "16": { | |
| "recovery_pct": 0.26, | |
| "logp_transplant_mean": -11.8098, | |
| "logp_transplant_std": 3.217 | |
| }, | |
| "17": { | |
| "recovery_pct": 0.01, | |
| "logp_transplant_mean": -11.8222, | |
| "logp_transplant_std": 3.2162 | |
| }, | |
| "18": { | |
| "recovery_pct": 0.16, | |
| "logp_transplant_mean": -11.8147, | |
| "logp_transplant_std": 3.151 | |
| }, | |
| "19": { | |
| "recovery_pct": 1.22, | |
| "logp_transplant_mean": -11.7631, | |
| "logp_transplant_std": 3.2305 | |
| }, | |
| "20": { | |
| "recovery_pct": 1.12, | |
| "logp_transplant_mean": -11.7679, | |
| "logp_transplant_std": 3.2154 | |
| }, | |
| "21": { | |
| "recovery_pct": 1.08, | |
| "logp_transplant_mean": -11.7699, | |
| "logp_transplant_std": 3.2113 | |
| }, | |
| "22": { | |
| "recovery_pct": 0.67, | |
| "logp_transplant_mean": -11.7898, | |
| "logp_transplant_std": 3.2088 | |
| }, | |
| "23": { | |
| "recovery_pct": 1.18, | |
| "logp_transplant_mean": -11.7651, | |
| "logp_transplant_std": 3.2582 | |
| }, | |
| "24": { | |
| "recovery_pct": 0.99, | |
| "logp_transplant_mean": -11.7742, | |
| "logp_transplant_std": 3.2714 | |
| }, | |
| "25": { | |
| "recovery_pct": 0.71, | |
| "logp_transplant_mean": -11.7881, | |
| "logp_transplant_std": 3.2825 | |
| }, | |
| "26": { | |
| "recovery_pct": 0.82, | |
| "logp_transplant_mean": -11.7824, | |
| "logp_transplant_std": 3.3644 | |
| }, | |
| "27": { | |
| "recovery_pct": 0.93, | |
| "logp_transplant_mean": -11.7772, | |
| "logp_transplant_std": 3.3586 | |
| }, | |
| "28": { | |
| "recovery_pct": 0.77, | |
| "logp_transplant_mean": -11.7851, | |
| "logp_transplant_std": 3.451 | |
| }, | |
| "29": { | |
| "recovery_pct": 2.33, | |
| "logp_transplant_mean": -11.7091, | |
| "logp_transplant_std": 3.3612 | |
| }, | |
| "30": { | |
| "recovery_pct": -1.05, | |
| "logp_transplant_mean": -11.8736, | |
| "logp_transplant_std": 3.2181 | |
| }, | |
| "31": { | |
| "recovery_pct": 21.97, | |
| "logp_transplant_mean": -10.751, | |
| "logp_transplant_std": 3.0641 | |
| }, | |
| "32": { | |
| "recovery_pct": 19.75, | |
| "logp_transplant_mean": -10.8595, | |
| "logp_transplant_std": 3.0857 | |
| }, | |
| "33": { | |
| "recovery_pct": 20.01, | |
| "logp_transplant_mean": -10.8466, | |
| "logp_transplant_std": 3.0286 | |
| }, | |
| "34": { | |
| "recovery_pct": 27.44, | |
| "logp_transplant_mean": -10.4844, | |
| "logp_transplant_std": 2.6999 | |
| }, | |
| "35": { | |
| "recovery_pct": 28.0, | |
| "logp_transplant_mean": -10.4571, | |
| "logp_transplant_std": 2.6612 | |
| }, | |
| "36": { | |
| "recovery_pct": 29.8, | |
| "logp_transplant_mean": -10.3692, | |
| "logp_transplant_std": 2.693 | |
| }, | |
| "37": { | |
| "recovery_pct": 30.63, | |
| "logp_transplant_mean": -10.329, | |
| "logp_transplant_std": 2.7026 | |
| }, | |
| "38": { | |
| "recovery_pct": 31.24, | |
| "logp_transplant_mean": -10.2992, | |
| "logp_transplant_std": 2.6515 | |
| }, | |
| "39": { | |
| "recovery_pct": 31.86, | |
| "logp_transplant_mean": -10.2687, | |
| "logp_transplant_std": 2.5616 | |
| }, | |
| "40": { | |
| "recovery_pct": 34.3, | |
| "logp_transplant_mean": -10.1501, | |
| "logp_transplant_std": 2.4601 | |
| }, | |
| "41": { | |
| "recovery_pct": 33.93, | |
| "logp_transplant_mean": -10.1681, | |
| "logp_transplant_std": 2.497 | |
| }, | |
| "42": { | |
| "recovery_pct": 73.91, | |
| "logp_transplant_mean": -8.2182, | |
| "logp_transplant_std": 3.4913 | |
| }, | |
| "43": { | |
| "recovery_pct": 75.3, | |
| "logp_transplant_mean": -8.1506, | |
| "logp_transplant_std": 3.5641 | |
| }, | |
| "44": { | |
| "recovery_pct": 76.57, | |
| "logp_transplant_mean": -8.0885, | |
| "logp_transplant_std": 3.5968 | |
| }, | |
| "45": { | |
| "recovery_pct": 106.31, | |
| "logp_transplant_mean": -6.6381, | |
| "logp_transplant_std": 3.475 | |
| }, | |
| "46": { | |
| "recovery_pct": 105.34, | |
| "logp_transplant_mean": -6.6857, | |
| "logp_transplant_std": 3.4952 | |
| }, | |
| "47": { | |
| "recovery_pct": 104.52, | |
| "logp_transplant_mean": -6.7256, | |
| "logp_transplant_std": 3.5292 | |
| }, | |
| "48": { | |
| "recovery_pct": 104.0, | |
| "logp_transplant_mean": -6.7509, | |
| "logp_transplant_std": 3.5304 | |
| }, | |
| "49": { | |
| "recovery_pct": 103.56, | |
| "logp_transplant_mean": -6.7726, | |
| "logp_transplant_std": 3.53 | |
| }, | |
| "50": { | |
| "recovery_pct": 103.53, | |
| "logp_transplant_mean": -6.7737, | |
| "logp_transplant_std": 3.5238 | |
| }, | |
| "51": { | |
| "recovery_pct": 103.62, | |
| "logp_transplant_mean": -6.7695, | |
| "logp_transplant_std": 3.5199 | |
| }, | |
| "52": { | |
| "recovery_pct": 103.19, | |
| "logp_transplant_mean": -6.7906, | |
| "logp_transplant_std": 3.5125 | |
| }, | |
| "53": { | |
| "recovery_pct": 103.89, | |
| "logp_transplant_mean": -6.7562, | |
| "logp_transplant_std": 3.511 | |
| }, | |
| "54": { | |
| "recovery_pct": 103.04, | |
| "logp_transplant_mean": -6.7979, | |
| "logp_transplant_std": 3.5106 | |
| }, | |
| "55": { | |
| "recovery_pct": 102.12, | |
| "logp_transplant_mean": -6.8425, | |
| "logp_transplant_std": 3.4931 | |
| }, | |
| "56": { | |
| "recovery_pct": 102.19, | |
| "logp_transplant_mean": -6.839, | |
| "logp_transplant_std": 3.4874 | |
| }, | |
| "57": { | |
| "recovery_pct": 102.32, | |
| "logp_transplant_mean": -6.8328, | |
| "logp_transplant_std": 3.4831 | |
| }, | |
| "58": { | |
| "recovery_pct": 102.18, | |
| "logp_transplant_mean": -6.8399, | |
| "logp_transplant_std": 3.4533 | |
| }, | |
| "59": { | |
| "recovery_pct": 102.06, | |
| "logp_transplant_mean": -6.8456, | |
| "logp_transplant_std": 3.4526 | |
| }, | |
| "60": { | |
| "recovery_pct": 101.65, | |
| "logp_transplant_mean": -6.8653, | |
| "logp_transplant_std": 3.4654 | |
| }, | |
| "61": { | |
| "recovery_pct": 101.48, | |
| "logp_transplant_mean": -6.8737, | |
| "logp_transplant_std": 3.4512 | |
| }, | |
| "62": { | |
| "recovery_pct": 101.09, | |
| "logp_transplant_mean": -6.8927, | |
| "logp_transplant_std": 3.4147 | |
| }, | |
| "63": { | |
| "recovery_pct": 101.34, | |
| "logp_transplant_mean": -6.8807, | |
| "logp_transplant_std": 3.3985 | |
| } | |
| }, | |
| "L_crit_90": 45, | |
| "L_crit_99": 45, | |
| "alpha_90": 0.7031, | |
| "alpha_99": 0.7031, | |
| "n_layers": 64, | |
| "d_model": 2560, | |
| "seed": 7, | |
| "runtime_seconds": 7521.5 | |
| } |