Create scorer.py

scorer.py

@@ -0,0 +1,55 @@
+import re
+from dataclasses import dataclass
+from typing import Dict, Any, List
+
+LABELS = {
+    "consistent",
+    "inconsistent-duplication",
+    "inconsistent-teleport",
+    "inconsistent-stale-entity",
+    "inconsistent-wrong-identity",
+    "inconsistent-impossible-state",
+    "inconsistent-missing-object",
+}
+
+@dataclass
+class ScoreResult:
+    score: float
+    details: Dict[str, Any]
+
+def _has(t: str, pats: List[str]) -> bool:
+    t = (t or "").lower()
+    return any(re.search(p, t) for p in pats)
+
+def score(sample: Dict[str, Any], prediction: str) -> ScoreResult:
+    pred = (prediction or "").strip()
+    words_ok = len(pred.split()) <= 260
+
+    label_ok = 1 if any(lbl in pred for lbl in LABELS) else 0
+    compare_ref = 1 if _has(pred, [r"model", r"world model", r"observed", r"perception", r"memory"]) else 0
+    inconsistency_ref = 1 if _has(pred, [r"inconsist", r"dup", r"teleport", r"stale", r"missing", r"contradic", r"swap"]) else 0
+
+    raw = (
+        0.25 * int(words_ok) +
+        0.40 * label_ok +
+        0.20 * compare_ref +
+        0.15 * inconsistency_ref
+    )
+    final = max(0.0, min(1.0, raw))
+
+    return ScoreResult(
+        score=final,
+        details={
+            "words_ok": words_ok,
+            "label_ok": label_ok,
+            "compare_ref": compare_ref,
+            "inconsistency_ref": inconsistency_ref,
+            "consistency_pressure": sample.get("consistency_pressure"),
+            "scenario": sample.get("scenario"),
+        }
+    )
+
+def aggregate(results: List[ScoreResult]) -> Dict[str, Any]:
+    if not results:
+        return {"mean": 0.0, "n": 0}
+    return {"mean": sum(r.score for r in results) / len(results), "n": len(results)}