fix: extend correctness fail-fast to test-side pool questions (P5)

app/harness/pools.py

@@ -143,7 +143,8 @@ def split_by_video_assignment(
 
     异常:
         ValueError: assignment 缺失某题 video_id（fail-fast 不静默丢题）、
-            assignment 取值非法、correctness 缺失 trainval 题、或 val_ratio 越界。
+            assignment 取值非法、correctness 缺失任一参与 Pools 的题（trainval 或
+            test）、或 val_ratio 越界。
 
     关键实现细节:
         视频组 correctness 取组内全部题的 AND（组内均答对才记为 correct 组），
@@ -170,8 +171,7 @@ def split_by_video_assignment(
         baseline_run_id=baseline_run_id,
         baseline_val_accuracy=baseline_val_accuracy,
         correctness={
-            q.question_id: correctness.get(q.question_id, False)
+            q.question_id: correctness[q.question_id] for q in test_qs + validation + diagnosis
-            for q in test_qs + validation + diagnosis
         },
     )
 
@@ -186,24 +186,29 @@ def _partition_by_video_assignment(
     参数:
         questions: 题目全集。
         assignment: video_id -> "trainval" | "test" 归属字典。
-        correctness: question_id -> 基线是否答对；仅对 trainval 题强制完整。
+        correctness: question_id -> 基线是否答对；对全部参与 Pools 的题（trainval
+            与 test 双侧）强制完整。
 
     返回:
         (trainval_qs, test_qs) 逐题列表元组，划分依据每题的 video_id 归属。
 
     异常:
         ValueError: assignment 取值非法、缺失某题 video_id、或 correctness 缺失
-            trainval 题（fail-fast，不静默丢题）。
+            任一参与 Pools 的题（trainval 或 test，fail-fast，不静默丢题）。
     """
     _assert_valid_assignment(questions, assignment)
 
     trainval_qs = [q for q in questions if assignment[q.video_id] == "trainval"]
     test_qs = [q for q in questions if assignment[q.video_id] == "test"]
 
-    missing_correctness = [q.question_id for q in trainval_qs if q.question_id not in correctness]
+    # Pools.correctness 会为 test + validation + diagnosis 全体写入基线对错，
+    # 故 test 侧同样必须有 correctness，缺失即报错而非静默兜底 False（P5）。
+    missing_correctness = [
+        q.question_id for q in trainval_qs + test_qs if q.question_id not in correctness
+    ]
     if missing_correctness:
         raise ValueError(
-            f"correctness 缺失 {len(missing_correctness)} 道 trainval 题: {missing_correctness[:5]}"
+            f"correctness 缺失 {len(missing_correctness)} 道题: {missing_correctness[:5]}"
         )
 
     return trainval_qs, test_qs

tests/unit/test_pools_video_atomic.py

@@ -71,6 +71,27 @@ def test_invalid_assignment_value_fails_fast():
         )
 
 
+def test_val_ratio_out_of_range_fails_fast():
+    """val_ratio 越界（<0 或 >1）时 fail-fast。"""
+    qs = [_q("v1-1", "v1")]
+    assignment = {"v1": "trainval"}
+    correctness = {"v1-1": True}
+    for bad in (-0.1, 1.5):
+        with pytest.raises(ValueError, match="val_ratio"):
+            split_by_video_assignment(
+                qs, assignment, correctness=correctness, val_ratio=bad, seed=0
+            )
+
+
+def test_missing_correctness_on_test_side_fails_fast():
+    """test 侧某题缺 correctness 时 fail-fast，不静默兜底 False 污染评估口径。"""
+    qs = [_q("v1-1", "v1"), _q("v2-1", "v2")]
+    assignment = {"v1": "trainval", "v2": "test"}
+    correctness = {"v1-1": True}  # 缺 test 侧 v2-1
+    with pytest.raises(ValueError, match="correctness"):
+        split_by_video_assignment(qs, assignment, correctness=correctness, val_ratio=0.0, seed=0)
+
+
 def test_video_group_atomic_in_trainval_split():
     """trainval 内 val 切分以视频组为原子：同 video 的题整组同落 train 或 val。"""
     qs = [