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": -6.5087, | |
| "std": 2.0768 | |
| }, | |
| "baseline_logp_d1000": { | |
| "mean": -11.7802, | |
| "std": 1.0988 | |
| }, | |
| "logp_gap": 5.2716, | |
| "layer_sweep": { | |
| "0": { | |
| "recovery_pct": 0.0, | |
| "logp_transplant_mean": -11.7802, | |
| "logp_transplant_std": 1.0988 | |
| }, | |
| "1": { | |
| "recovery_pct": -0.18, | |
| "logp_transplant_mean": -11.7898, | |
| "logp_transplant_std": 1.1569 | |
| }, | |
| "2": { | |
| "recovery_pct": 4.28, | |
| "logp_transplant_mean": -11.5544, | |
| "logp_transplant_std": 1.1592 | |
| }, | |
| "3": { | |
| "recovery_pct": 3.96, | |
| "logp_transplant_mean": -11.5714, | |
| "logp_transplant_std": 1.1857 | |
| }, | |
| "4": { | |
| "recovery_pct": 3.71, | |
| "logp_transplant_mean": -11.5848, | |
| "logp_transplant_std": 1.1628 | |
| }, | |
| "5": { | |
| "recovery_pct": 4.01, | |
| "logp_transplant_mean": -11.5686, | |
| "logp_transplant_std": 1.2155 | |
| }, | |
| "6": { | |
| "recovery_pct": 4.07, | |
| "logp_transplant_mean": -11.5657, | |
| "logp_transplant_std": 1.1764 | |
| }, | |
| "7": { | |
| "recovery_pct": 4.22, | |
| "logp_transplant_mean": -11.5579, | |
| "logp_transplant_std": 1.1926 | |
| }, | |
| "8": { | |
| "recovery_pct": 3.86, | |
| "logp_transplant_mean": -11.5767, | |
| "logp_transplant_std": 1.1623 | |
| }, | |
| "9": { | |
| "recovery_pct": 3.97, | |
| "logp_transplant_mean": -11.571, | |
| "logp_transplant_std": 1.1581 | |
| }, | |
| "10": { | |
| "recovery_pct": 4.03, | |
| "logp_transplant_mean": -11.5679, | |
| "logp_transplant_std": 1.1707 | |
| }, | |
| "11": { | |
| "recovery_pct": 7.93, | |
| "logp_transplant_mean": -11.3621, | |
| "logp_transplant_std": 1.2879 | |
| }, | |
| "12": { | |
| "recovery_pct": 7.24, | |
| "logp_transplant_mean": -11.3985, | |
| "logp_transplant_std": 1.2844 | |
| }, | |
| "13": { | |
| "recovery_pct": 7.22, | |
| "logp_transplant_mean": -11.3995, | |
| "logp_transplant_std": 1.2797 | |
| }, | |
| "14": { | |
| "recovery_pct": 7.0, | |
| "logp_transplant_mean": -11.4112, | |
| "logp_transplant_std": 1.2606 | |
| }, | |
| "15": { | |
| "recovery_pct": 7.11, | |
| "logp_transplant_mean": -11.4054, | |
| "logp_transplant_std": 1.2606 | |
| }, | |
| "16": { | |
| "recovery_pct": 6.51, | |
| "logp_transplant_mean": -11.4371, | |
| "logp_transplant_std": 1.2229 | |
| }, | |
| "17": { | |
| "recovery_pct": 6.77, | |
| "logp_transplant_mean": -11.4235, | |
| "logp_transplant_std": 1.2163 | |
| }, | |
| "18": { | |
| "recovery_pct": 6.26, | |
| "logp_transplant_mean": -11.4501, | |
| "logp_transplant_std": 1.209 | |
| }, | |
| "19": { | |
| "recovery_pct": 7.34, | |
| "logp_transplant_mean": -11.3933, | |
| "logp_transplant_std": 1.2653 | |
| }, | |
| "20": { | |
| "recovery_pct": 7.19, | |
| "logp_transplant_mean": -11.401, | |
| "logp_transplant_std": 1.264 | |
| }, | |
| "21": { | |
| "recovery_pct": 7.37, | |
| "logp_transplant_mean": -11.3919, | |
| "logp_transplant_std": 1.2945 | |
| }, | |
| "22": { | |
| "recovery_pct": 7.25, | |
| "logp_transplant_mean": -11.398, | |
| "logp_transplant_std": 1.3094 | |
| }, | |
| "23": { | |
| "recovery_pct": 7.35, | |
| "logp_transplant_mean": -11.3928, | |
| "logp_transplant_std": 1.3744 | |
| }, | |
| "24": { | |
| "recovery_pct": 8.83, | |
| "logp_transplant_mean": -11.3146, | |
| "logp_transplant_std": 1.5151 | |
| }, | |
| "25": { | |
| "recovery_pct": 9.48, | |
| "logp_transplant_mean": -11.2802, | |
| "logp_transplant_std": 1.5697 | |
| }, | |
| "26": { | |
| "recovery_pct": 9.98, | |
| "logp_transplant_mean": -11.2541, | |
| "logp_transplant_std": 1.5491 | |
| }, | |
| "27": { | |
| "recovery_pct": 10.02, | |
| "logp_transplant_mean": -11.2519, | |
| "logp_transplant_std": 1.5208 | |
| }, | |
| "28": { | |
| "recovery_pct": 9.53, | |
| "logp_transplant_mean": -11.2776, | |
| "logp_transplant_std": 1.5642 | |
| }, | |
| "29": { | |
| "recovery_pct": 12.36, | |
| "logp_transplant_mean": -11.1287, | |
| "logp_transplant_std": 1.6826 | |
| }, | |
| "30": { | |
| "recovery_pct": 8.2, | |
| "logp_transplant_mean": -11.3479, | |
| "logp_transplant_std": 1.4459 | |
| }, | |
| "31": { | |
| "recovery_pct": 26.2, | |
| "logp_transplant_mean": -10.3988, | |
| "logp_transplant_std": 1.8647 | |
| }, | |
| "32": { | |
| "recovery_pct": 24.93, | |
| "logp_transplant_mean": -10.4661, | |
| "logp_transplant_std": 1.8088 | |
| }, | |
| "33": { | |
| "recovery_pct": 25.86, | |
| "logp_transplant_mean": -10.4168, | |
| "logp_transplant_std": 1.8057 | |
| }, | |
| "34": { | |
| "recovery_pct": 29.76, | |
| "logp_transplant_mean": -10.2111, | |
| "logp_transplant_std": 1.8414 | |
| }, | |
| "35": { | |
| "recovery_pct": 30.15, | |
| "logp_transplant_mean": -10.1907, | |
| "logp_transplant_std": 1.8758 | |
| }, | |
| "36": { | |
| "recovery_pct": 32.33, | |
| "logp_transplant_mean": -10.0758, | |
| "logp_transplant_std": 1.9394 | |
| }, | |
| "37": { | |
| "recovery_pct": 33.41, | |
| "logp_transplant_mean": -10.0191, | |
| "logp_transplant_std": 1.933 | |
| }, | |
| "38": { | |
| "recovery_pct": 33.89, | |
| "logp_transplant_mean": -9.9936, | |
| "logp_transplant_std": 1.9287 | |
| }, | |
| "39": { | |
| "recovery_pct": 34.85, | |
| "logp_transplant_mean": -9.9429, | |
| "logp_transplant_std": 1.9033 | |
| }, | |
| "40": { | |
| "recovery_pct": 35.38, | |
| "logp_transplant_mean": -9.915, | |
| "logp_transplant_std": 1.9584 | |
| }, | |
| "41": { | |
| "recovery_pct": 33.84, | |
| "logp_transplant_mean": -9.9961, | |
| "logp_transplant_std": 1.8755 | |
| }, | |
| "42": { | |
| "recovery_pct": 76.42, | |
| "logp_transplant_mean": -7.7516, | |
| "logp_transplant_std": 2.2565 | |
| }, | |
| "43": { | |
| "recovery_pct": 78.8, | |
| "logp_transplant_mean": -7.6264, | |
| "logp_transplant_std": 2.3226 | |
| }, | |
| "44": { | |
| "recovery_pct": 76.55, | |
| "logp_transplant_mean": -7.7451, | |
| "logp_transplant_std": 2.2208 | |
| }, | |
| "45": { | |
| "recovery_pct": 108.26, | |
| "logp_transplant_mean": -6.0735, | |
| "logp_transplant_std": 2.0031 | |
| }, | |
| "46": { | |
| "recovery_pct": 107.72, | |
| "logp_transplant_mean": -6.1015, | |
| "logp_transplant_std": 2.0121 | |
| }, | |
| "47": { | |
| "recovery_pct": 106.7, | |
| "logp_transplant_mean": -6.1557, | |
| "logp_transplant_std": 2.0251 | |
| }, | |
| "48": { | |
| "recovery_pct": 106.67, | |
| "logp_transplant_mean": -6.1569, | |
| "logp_transplant_std": 2.0271 | |
| }, | |
| "49": { | |
| "recovery_pct": 106.41, | |
| "logp_transplant_mean": -6.1706, | |
| "logp_transplant_std": 2.0596 | |
| }, | |
| "50": { | |
| "recovery_pct": 106.28, | |
| "logp_transplant_mean": -6.1775, | |
| "logp_transplant_std": 2.0684 | |
| }, | |
| "51": { | |
| "recovery_pct": 105.52, | |
| "logp_transplant_mean": -6.2176, | |
| "logp_transplant_std": 2.0735 | |
| }, | |
| "52": { | |
| "recovery_pct": 104.77, | |
| "logp_transplant_mean": -6.2573, | |
| "logp_transplant_std": 2.1065 | |
| }, | |
| "53": { | |
| "recovery_pct": 105.33, | |
| "logp_transplant_mean": -6.2278, | |
| "logp_transplant_std": 2.1274 | |
| }, | |
| "54": { | |
| "recovery_pct": 103.64, | |
| "logp_transplant_mean": -6.3166, | |
| "logp_transplant_std": 2.1361 | |
| }, | |
| "55": { | |
| "recovery_pct": 102.62, | |
| "logp_transplant_mean": -6.3704, | |
| "logp_transplant_std": 2.1232 | |
| }, | |
| "56": { | |
| "recovery_pct": 102.16, | |
| "logp_transplant_mean": -6.3951, | |
| "logp_transplant_std": 2.1115 | |
| }, | |
| "57": { | |
| "recovery_pct": 102.05, | |
| "logp_transplant_mean": -6.4004, | |
| "logp_transplant_std": 2.1216 | |
| }, | |
| "58": { | |
| "recovery_pct": 101.97, | |
| "logp_transplant_mean": -6.4048, | |
| "logp_transplant_std": 2.0898 | |
| }, | |
| "59": { | |
| "recovery_pct": 101.38, | |
| "logp_transplant_mean": -6.4357, | |
| "logp_transplant_std": 2.095 | |
| }, | |
| "60": { | |
| "recovery_pct": 101.39, | |
| "logp_transplant_mean": -6.4352, | |
| "logp_transplant_std": 2.0981 | |
| }, | |
| "61": { | |
| "recovery_pct": 101.21, | |
| "logp_transplant_mean": -6.4447, | |
| "logp_transplant_std": 2.0842 | |
| }, | |
| "62": { | |
| "recovery_pct": 100.34, | |
| "logp_transplant_mean": -6.4907, | |
| "logp_transplant_std": 2.0915 | |
| }, | |
| "63": { | |
| "recovery_pct": 100.91, | |
| "logp_transplant_mean": -6.4606, | |
| "logp_transplant_std": 2.0897 | |
| } | |
| }, | |
| "L_crit_90": 45, | |
| "L_crit_99": 45, | |
| "alpha_90": 0.7031, | |
| "alpha_99": 0.7031, | |
| "n_layers": 64, | |
| "d_model": 2560, | |
| "seed": 123, | |
| "runtime_seconds": 8278.2 | |
| } |