初始化项目，由ModelHub XC社区提供模型

Model: Zynerji/Ektome-SmolLM2-360Mi-PristinelyUncensored Source: Original Platform
2026-09-19 05:22:17 +08:00
commit 09886af51e
27 changed files with 490517 additions and 0 deletions
--- a/cert_SmolLM2-360Mi.json
+++ b/cert_SmolLM2-360Mi.json
@@ -0,0 +1,133 @@
+{
+  "sphragis_version": "0.1.0",
+  "generated_at": "2026-07-27T16:17:13.291817+00:00",
+  "reference": {
+    "endpoint": "http://127.0.0.1:8080/v1",
+    "model": "ref"
+  },
+  "candidate": {
+    "endpoint": "http://127.0.0.1:8081/v1",
+    "model": "cand"
+  },
+  "pack": {
+    "name": "/root/pack_v2.jsonl",
+    "n_tasks": 2800,
+    "sha256": "2de27099bbb15bab4f7b35599b215f038fdb68b3f5f4e1cda1a464f9dd18e14e"
+  },
+  "overall": "FAIL",
+  "claim": "At least one axis shows a statistically significant accuracy regression (exact McNemar, Holm-corrected, alpha=0.05).",
+  "axes": [
+    {
+      "axis": "arithmetic",
+      "n": 1400,
+      "counts": {
+        "both_correct": 237,
+        "ref_only": 203,
+        "cand_only": 88,
+        "both_wrong": 872
+      },
+      "acc_reference": 0.314286,
+      "acc_candidate": 0.232143,
+      "regression_d": 0.082143,
+      "d_ci": [
+        0.058571,
+        0.106429
+      ],
+      "d_upper_bound": 0.102857,
+      "p_regression": 0.0,
+      "p_regression_holm": 0.0,
+      "p_improvement": 1.0,
+      "improved": false,
+      "mde_at_power": 0.030275,
+      "n_needed_for_margin": 1426,
+      "verdict": "FAIL",
+      "reason": "significant_regression"
+    },
+    {
+      "axis": "instruction",
+      "n": 600,
+      "counts": {
+        "both_correct": 0,
+        "ref_only": 11,
+        "cand_only": 16,
+        "both_wrong": 573
+      },
+      "acc_reference": 0.018333,
+      "acc_candidate": 0.026667,
+      "regression_d": -0.008333,
+      "d_ci": [
+        -0.025,
+        0.008333
+      ],
+      "d_upper_bound": 0.005,
+      "p_regression": 0.87610572,
+      "p_regression_holm": 1.0,
+      "p_improvement": 0.22103417,
+      "improved": false,
+      "mde_at_power": 0.021496,
+      "n_needed_for_margin": 308,
+      "verdict": "PASS",
+      "reason": "non_inferior_within_margin"
+    },
+    {
+      "axis": "knowledge",
+      "n": 400,
+      "counts": {
+        "both_correct": 57,
+        "ref_only": 67,
+        "cand_only": 38,
+        "both_wrong": 238
+      },
+      "acc_reference": 0.31,
+      "acc_candidate": 0.2375,
+      "regression_d": 0.0725,
+      "d_ci": [
+        0.0225,
+        0.1225
+      ],
+      "d_upper_bound": 0.115,
+      "p_regression": 0.00300804,
+      "p_regression_holm": 0.00902412,
+      "p_improvement": 0.99838951,
+      "improved": false,
+      "mde_at_power": 0.063531,
+      "n_needed_for_margin": 1802,
+      "verdict": "FAIL",
+      "reason": "significant_regression"
+    },
+    {
+      "axis": "reasoning",
+      "n": 400,
+      "counts": {
+        "both_correct": 50,
+        "ref_only": 34,
+        "cand_only": 35,
+        "both_wrong": 281
+      },
+      "acc_reference": 0.21,
+      "acc_candidate": 0.2125,
+      "regression_d": -0.0025,
+      "d_ci": [
+        -0.045,
+        0.0375
+      ],
+      "d_upper_bound": 0.0325,
+      "p_regression": 0.59502547,
+      "p_regression_holm": 1.0,
+      "p_improvement": 0.5,
+      "improved": false,
+      "mde_at_power": 0.051501,
+      "n_needed_for_margin": 1183,
+      "verdict": "INCONCLUSIVE",
+      "reason": "ci_too_wide"
+    }
+  ],
+  "params": {
+    "margin": 0.03,
+    "alpha": 0.05,
+    "n_floor": 30,
+    "power": 0.8,
+    "n_boot": 4000,
+    "seed": 0
+  }
+}