fix: attribute descriptive review gaps to the named evidence target

backend/tests/test_optimization.py

@@ -373,6 +373,34 @@ def test_model_declining_score_records_original_gap(sqlite_db):
     assert score.evidence_gaps[0]["reason"] == "只有装置介绍"
 
 
+@pytest.mark.parametrize("missing", [["误差分析（t1）：缺少误差来源和统计数据"],
+                                      ["尚缺 t1 对应的正文"], ["当前资料无法判断"]])
+def test_declined_score_attributes_descriptive_missing_evidence_to_its_target(sqlite_db, missing):
+    db = sqlite_db
+    item = project(db)
+    file = SubmissionFile(project_id=item.id, category="stage3", relative_path="test.pdf", mime_type="application/pdf",
+                          size=1, mtime=1, sha256="test", recognition_status="recognized", parse_status="ready")
+    db.add(file)
+    db.commit()
+    ingest(db, course_version_id=1, project_id=item.id, submission_file_id=file.id, title="test.pdf",
+           category="stage3", source_hash="test", pages=[{"page_number": 1, "text": "实验测得响应时间80毫秒，但未解释误差来源。"}], embedder=FLAT)
+    criterion = {"criterion_id": "stage3.data", "category": "stage3", "max_score": 20,
+                 "description": "实验数据与误差分析", "required_evidence": ["实验数据", "误差分析"]}
+    created = services.create_batch(SimpleNamespace(course_version_id=1, rubric_version_id=1), db)
+    class Client:
+        def json_call(self, *args, **kwargs):
+            return ScoreResult(criterion_id="stage3.data", score=None, max_score=20,
+                               reason="误差分析尚不足以判断", missing_evidence=missing, confidence=.2)
+    def retriever(session, **kwargs):
+        if kwargs.get("public_only"):
+            return []
+        return retrieve(session, **{**kwargs, "embedder": FLAT})
+    score = _score_criterion(db, db.get(ScoringBatch, created["id"]), item, criterion, Client(), retriever=retriever)
+    expected = ["t0", "t1"] if missing == ["当前资料无法判断"] else ["t1"]
+    assert [gap["target_id"] for gap in score.evidence_gaps] == expected
+    assert score.status == "needs_review" and score.model_score is None
+
+
 def test_failed_model_call_cannot_overwrite_a_concurrent_teacher_review(sqlite_db, monkeypatch):
     from wisegrade import scoring
     db = sqlite_db

backend/wisegrade/scoring.py

@@ -1,4 +1,5 @@
 from concurrent.futures import ThreadPoolExecutor, as_completed
+import re
 
 from sqlalchemy import and_, or_, select, update
 
@@ -258,8 +259,13 @@ def _score_criterion(db, batch, project, criterion, client, *, retriever, multi_
     if result.score is None:
         # The retrieval found body pages but the model judged them insufficient.
         score.status, score.error = STATUS_NEEDS_REVIEW, "模型认为现有正文证据不足以判断该项，转教师复核。"
+        # Models normally write "误差分析（t1）：缺少…", not just the bare
+        # label. Attribute those explanations to their target instead of marking
+        # every target as missing. Unknown explanations remain conservative.
         declined = [target for target in targets if target["use"] == USE_CONTENT and
-                    (not result.missing_evidence or target["label"] in result.missing_evidence)]
+                    any(target["label"] in note or re.search(
+                        r"(?<![A-Za-z0-9_])" + re.escape(target["target_id"]) + r"(?![A-Za-z0-9_])", note)
+                        for note in result.missing_evidence)]
         _record_evidence_gaps(score, declined or [target for target in targets if target["use"] == USE_CONTENT])
         for gap in score.evidence_gaps:
             gap["reason"] = result.reason