[ { "model": "hermes3:8b", "params": "8.0B", "quant": "Q4_0", "digest": "4f6b83f30b62", "ollama": "0.18.2", "status": "completed", "games": 10, "wins": 0, "win_ci_low": 0.0, "win_ci_high": 0.2775, "treasure": 0, "outcomes": { "death_3hits": 2, "timeout": 8 }, "mean_rounds": 120.8, "llm_calls": 1258, "invalid": 0, "invalid_by_reason": {}, "forced_passes": 0, "timed_out_calls": 0, "timeout_rate": 0.0, "perturbed": false, "warmup": { "status": "ok", "duration_ms": 9402, "counted": false }, "mean_latency_ms": 1004.0, "known_wall_repeats": 1188, "mean_bumps": 120.6, "samples": 1258, "mean_conf": 0.7999, "move_success": 0.0413, "move_ci_low": 0.0317, "move_ci_high": 0.0538, "brier": 0.6152, "ece": 0.7591, "ece_ci_low": 0.7479, "ece_ci_high": 0.7698, "resolution": 0.0007, "reliability": 0.5763, "uncertainty": 0.0396, "bins": [ { "lower": 0.7, "upper": 0.8, "samples": 1, "mean_confidence": 0.7, "observed_rate": 1.0, "observed_rate_ci95_low": 0.2065, "observed_rate_ci95_high": 1.0, "sufficient": false }, { "lower": 0.8, "upper": 0.9, "samples": 1257, "mean_confidence": 0.8, "observed_rate": 0.0406, "observed_rate_ci95_low": 0.031, "observed_rate_ci95_high": 0.053, "sufficient": true } ], "plain": "Quand le modèle se dit sûr à environ 80 %, il a raison 4 % du temps (sur 1257 coups ; plage plausible 3 % à 5 %). En moyenne, il annonce 80 % et réussit 4 % de ses déplacements.", "wall_seconds": 1274, "report_sha256": "9a739479938cf89925f9479eab0ea59fa51736fbd77e3084e3f39f7fa7e6edc7", "evidence": [ "62e8eb5ba912", "fb416901283e", "a8e97c2a8ccb", "f2f464d27171", "aa37d062ef3d", "476b82a07059", "0afaff6d2d90", "2ec0efaf5f77", "5b147aea0268", "2c8ced246799" ], "game_outcomes": [ "death_3hits", "timeout", "death_3hits", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout" ] }, { "model": "mistral-nemo:12b", "params": "12.2B", "quant": "Q4_0", "digest": "e7e06d107c6c", "ollama": "0.18.2", "status": "completed", "games": 10, "wins": 0, "win_ci_low": 0.0, "win_ci_high": 0.2775, "treasure": 0, "outcomes": { "timeout": 10 }, "mean_rounds": 142.8, "llm_calls": 1991, "invalid": 0, "invalid_by_reason": {}, "forced_passes": 0, "timed_out_calls": 0, "timeout_rate": 0.0, "perturbed": false, "warmup": { "status": "ok", "duration_ms": 10507, "counted": false }, "mean_latency_ms": 1299.0, "known_wall_repeats": 1383, "mean_bumps": 142.2, "samples": 1991, "mean_conf": 0.9484, "move_success": 0.2858, "move_ci_low": 0.2664, "move_ci_high": 0.306, "brier": 0.6426, "ece": 0.6626, "ece_ci_low": 0.6408, "ece_ci_high": 0.6816, "resolution": 0.0, "reliability": 0.4391, "uncertainty": 0.2041, "bins": [ { "lower": 0.8, "upper": 0.9, "samples": 2, "mean_confidence": 0.85, "observed_rate": 0.5, "observed_rate_ci95_low": 0.0945, "observed_rate_ci95_high": 0.9055, "sufficient": false }, { "lower": 0.9, "upper": 1.0, "samples": 1989, "mean_confidence": 0.9485, "observed_rate": 0.2856, "observed_rate_ci95_low": 0.2661, "observed_rate_ci95_high": 0.3058, "sufficient": true } ], "plain": "Quand le modèle se dit sûr à environ 95 %, il a raison 29 % du temps (sur 1989 coups ; plage plausible 27 % à 31 %). En moyenne, il annonce 95 % et réussit 29 % de ses déplacements.", "wall_seconds": 2599, "report_sha256": "36df2ea08813aa25ee243269d5cfe5e2df8a4a399e7e6062a53434184f227154", "evidence": [ "a669b32b33a3", "b24ad1352352", "2fadb381bd5e", "76e2f01ad311", "1feaba143c53", "ab143e768f18", "0a4a530fab26", "f4f8029e0a4f", "92997544584e", "5d9839ca4d37" ], "game_outcomes": [ "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout" ] }, { "model": "mistral:latest", "params": "7.2B", "quant": "Q4_K_M", "digest": "6577803aa9a0", "ollama": "0.18.2", "status": "completed", "games": 10, "wins": 0, "win_ci_low": 0.0, "win_ci_high": 0.2775, "treasure": 0, "outcomes": { "timeout": 10 }, "mean_rounds": 150.0, "llm_calls": 1778, "invalid": 0, "invalid_by_reason": {}, "forced_passes": 0, "timed_out_calls": 0, "timeout_rate": 0.0, "perturbed": false, "warmup": { "status": "ok", "duration_ms": 6934, "counted": false }, "mean_latency_ms": 583.0, "known_wall_repeats": 634, "mean_bumps": 64.5, "samples": 957, "mean_conf": 0.8492, "move_success": 0.326, "move_ci_low": 0.2971, "move_ci_high": 0.3564, "brier": 0.5061, "ece": 0.541, "ece_ci_low": 0.5116, "ece_ci_high": 0.5692, "resolution": 0.0132, "reliability": 0.3029, "uncertainty": 0.2197, "bins": [ { "lower": 0.5, "upper": 0.6, "samples": 55, "mean_confidence": 0.5, "observed_rate": 0.6545, "observed_rate_ci95_low": 0.5225, "observed_rate_ci95_high": 0.7664, "sufficient": true }, { "lower": 0.8, "upper": 0.9, "samples": 296, "mean_confidence": 0.8, "observed_rate": 0.1858, "observed_rate_ci95_low": 0.1456, "observed_rate_ci95_high": 0.234, "sufficient": true }, { "lower": 0.9, "upper": 1.0, "samples": 606, "mean_confidence": 0.905, "observed_rate": 0.3647, "observed_rate_ci95_low": 0.3273, "observed_rate_ci95_high": 0.4038, "sufficient": true } ], "plain": "Quand le modèle se dit sûr à environ 91 %, il a raison 36 % du temps (sur 606 coups ; plage plausible 33 % à 40 %). En moyenne, il annonce 85 % et réussit 33 % de ses déplacements.", "wall_seconds": 1045, "report_sha256": "64be3a258afdfdeca0c562093efe78e859fadaefa00c60c0312306db43778c59", "evidence": [ "ae1c3544c000", "a10e7b6fe06c", "61cff7cf4e6d", "ac503fbdf3f6", "9465fd989a5b", "84663d5fe8a7", "0ff2e57d211d", "5094483b20a6", "ad8e8a130913", "0b1846813fb7" ], "game_outcomes": [ "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout" ] }, { "model": "qwen2.5:14b", "params": "14.8B", "quant": "Q4_K_M", "digest": "7cdf5a0187d5", "ollama": "0.18.2", "status": "completed", "games": 10, "wins": 0, "win_ci_low": 0.0, "win_ci_high": 0.2775, "treasure": 0, "outcomes": { "death_3hits": 3, "timeout": 7 }, "mean_rounds": 96.7, "llm_calls": 1497, "invalid": 0, "invalid_by_reason": {}, "forced_passes": 0, "timed_out_calls": 0, "timeout_rate": 0.0, "perturbed": false, "warmup": { "status": "ok", "duration_ms": 12961, "counted": false }, "mean_latency_ms": 1311.0, "known_wall_repeats": 884, "mean_bumps": 91.1, "samples": 1497, "mean_conf": 0.748, "move_success": 0.3914, "move_ci_low": 0.367, "move_ci_high": 0.4164, "brier": 0.352, "ece": 0.3565, "ece_ci_low": 0.3329, "ece_ci_high": 0.3801, "resolution": 0.0217, "reliability": 0.1356, "uncertainty": 0.2382, "bins": [ { "lower": 0.5, "upper": 0.6, "samples": 165, "mean_confidence": 0.5, "observed_rate": 0.0121, "observed_rate_ci95_low": 0.0033, "observed_rate_ci95_high": 0.0431, "sufficient": true }, { "lower": 0.7, "upper": 0.8, "samples": 1032, "mean_confidence": 0.75, "observed_rate": 0.4496, "observed_rate_ci95_low": 0.4195, "observed_rate_ci95_high": 0.4801, "sufficient": true }, { "lower": 0.8, "upper": 0.9, "samples": 223, "mean_confidence": 0.85, "observed_rate": 0.3229, "observed_rate_ci95_low": 0.2649, "observed_rate_ci95_high": 0.3868, "sufficient": true }, { "lower": 0.9, "upper": 1.0, "samples": 77, "mean_confidence": 0.9571, "observed_rate": 0.6234, "observed_rate_ci95_low": 0.5117, "observed_rate_ci95_high": 0.7233, "sufficient": true } ], "plain": "Quand le modèle se dit sûr à environ 75 %, il a raison 45 % du temps (sur 1032 coups ; plage plausible 42 % à 48 %). En moyenne, il annonce 75 % et réussit 39 % de ses déplacements.", "wall_seconds": 1977, "report_sha256": "17819f935c0631a60f8cdc373557c7bffc1ad061f5597a20488bfee842a4af12", "evidence": [ "cf514245fbfa", "afacd82dca62", "943af64fc95c", "d237e64097b8", "e806a318a10a", "29ca38970472", "886a59318745", "f05aee58f872", "5f149e55f345", "9b78e1fc6feb" ], "game_outcomes": [ "timeout", "timeout", "death_3hits", "timeout", "timeout", "timeout", "timeout", "death_3hits", "death_3hits", "timeout" ] }, { "model": "qwen2.5:latest", "params": "7.6B", "quant": "Q4_K_M", "digest": "845dbda0ea48", "ollama": "0.18.2", "status": "completed", "games": 10, "wins": 0, "win_ci_low": 0.0, "win_ci_high": 0.2775, "treasure": 0, "outcomes": { "death_3hits": 1, "timeout": 9 }, "mean_rounds": 135.4, "llm_calls": 1395, "invalid": 0, "invalid_by_reason": {}, "forced_passes": 0, "timed_out_calls": 0, "timeout_rate": 0.0, "perturbed": false, "warmup": { "status": "ok", "duration_ms": 10026, "counted": false }, "mean_latency_ms": 1244.0, "known_wall_repeats": 1343, "mean_bumps": 135.3, "samples": 1395, "mean_conf": 0.9531, "move_success": 0.0301, "move_ci_low": 0.0224, "move_ci_high": 0.0404, "brier": 0.8804, "ece": 0.923, "ece_ci_low": 0.9137, "ece_ci_high": 0.9314, "resolution": 0.0, "reliability": 0.8519, "uncertainty": 0.0292, "bins": [ { "lower": 0.9, "upper": 1.0, "samples": 1395, "mean_confidence": 0.9531, "observed_rate": 0.0301, "observed_rate_ci95_low": 0.0224, "observed_rate_ci95_high": 0.0404, "sufficient": true } ], "plain": "Quand le modèle se dit sûr à environ 95 %, il a raison 3 % du temps (sur 1395 coups ; plage plausible 2 % à 4 %). En moyenne, il annonce 95 % et réussit 3 % de ses déplacements.", "wall_seconds": 1747, "report_sha256": "c837922ac1b5ff0d51f5ed2d9c535315a969875ae71c4044a474a979b79227c1", "evidence": [ "92ebf03c9abf", "a9ec305d6267", "1192909c2665", "bc412bf5dcd4", "5dcdd76a0fd8", "610a570574a7", "10f22d52d24c", "26b83ff825f7", "a83c0493ebce", "a537ae3445c3" ], "game_outcomes": [ "timeout", "timeout", "death_3hits", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout" ] }, { "model": "qwen3:0.6b", "params": "751.63M", "quant": "Q4_K_M", "digest": "7df6b6e09427", "ollama": "0.18.2", "status": "completed", "games": 10, "wins": 0, "win_ci_low": 0.0, "win_ci_high": 0.2775, "treasure": 0, "outcomes": { "timeout": 10 }, "mean_rounds": 150.0, "llm_calls": 1500, "invalid": 0, "invalid_by_reason": {}, "forced_passes": 0, "timed_out_calls": 0, "timeout_rate": 0.0, "perturbed": false, "warmup": { "status": "ok", "duration_ms": 2920, "counted": false }, "mean_latency_ms": 478.0, "known_wall_repeats": 208, "mean_bumps": 21.8, "samples": 218, "mean_conf": 0.8392, "move_success": 0.0, "move_ci_low": 0.0, "move_ci_high": 0.0173, "brier": 0.7747, "ece": 0.8392, "ece_ci_low": 0.8036, "ece_ci_high": 0.8738, "resolution": 0.0, "reliability": 0.7728, "uncertainty": 0.0, "bins": [ { "lower": 0.0, "upper": 0.1, "samples": 17, "mean_confidence": 0.0, "observed_rate": 0.0, "observed_rate_ci95_low": 0.0, "observed_rate_ci95_high": 0.1843, "sufficient": true }, { "lower": 0.5, "upper": 0.6, "samples": 11, "mean_confidence": 0.5, "observed_rate": 0.0, "observed_rate_ci95_low": 0.0, "observed_rate_ci95_high": 0.2588, "sufficient": true }, { "lower": 0.9, "upper": 1.0, "samples": 190, "mean_confidence": 0.9339, "observed_rate": 0.0, "observed_rate_ci95_low": 0.0, "observed_rate_ci95_high": 0.0198, "sufficient": true } ], "plain": "Quand le modèle se dit sûr à environ 93 %, il a raison 0 % du temps (sur 190 coups ; plage plausible 0 % à 2 %). En moyenne, il annonce 84 % et réussit 0 % de ses déplacements.", "wall_seconds": 721, "report_sha256": "dc251b237a6b96530e4d70d01aee8ccbaa55e533d8f9c16af8f402282dbfc6f4", "evidence": [ "367cd9120b98", "709909bee892", "6fdb1ca59d11", "b782cb86c42d", "7506c2d3167d", "c76143e24316", "4557923ba7b3", "f88f71a83bd2", "540c290097b1", "a7e2126eb6a2" ], "game_outcomes": [ "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout", "timeout" ] } ]