feat(tools): generate_questions.py calibrate 子命令

- Fisher exact test + effect size 组合判定（PASS/WARN/FAIL）
- 按 video_id 分组推理，避免跨视频树错用
- baseline 支持从 DB 读取或自动跑推理
- 对比表输出 + 退出码控制

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

tests/unit/test_generate_questions.py

@@ -11,6 +11,7 @@ import sys
 from pathlib import Path
 
 import numpy as np
+import pytest
 
 # 确保项目根目录在 sys.path 中
 PROJECT_ROOT = Path(__file__).resolve().parent.parent.parent
@@ -20,10 +21,13 @@ if str(PROJECT_ROOT) not in sys.path:
 from core.types import GeneratedQuestion
 from tools.generate_questions import (
     _append_to_json,
+    _calibrate_exit_code,
+    _judge_task_type,
     _load_or_init_progress,
     _rebuild_embedding_pool,
     _save_progress,
     _select_exemplars,
+    _validate_calibrate_args,
 )
 
 # ---------------------------------------------------------------------------
@@ -320,3 +324,94 @@ class TestAppendToJson:
         v2_data = json.loads((tmp_path / "v2.json").read_text())
         assert len(v1_data) == 1
         assert len(v2_data) == 1
+
+
+# ---------------------------------------------------------------------------
+# TestCalibrateJudgment
+# ---------------------------------------------------------------------------
+
+
+class TestCalibrateJudgment:
+    """_judge_task_type 校准判定测试。"""
+
+    def test_pass_when_delta_small(self) -> None:
+        """差值在容忍范围内判定为 PASS。"""
+        verdict = _judge_task_type(
+            bench_correct=60,
+            bench_total=100,
+            gen_correct=12,
+            gen_total=20,
+            tolerance=0.10,
+            alpha=0.05,
+        )
+        assert verdict == "PASS"
+
+    def test_fail_when_delta_large_and_significant(self) -> None:
+        """差值超阈值且统计显著判定为 FAIL。"""
+        verdict = _judge_task_type(
+            bench_correct=144,
+            bench_total=240,
+            gen_correct=6,
+            gen_total=20,
+            tolerance=0.10,
+            alpha=0.05,
+        )
+        assert verdict == "FAIL"
+
+    def test_warn_when_delta_large_but_not_significant(self) -> None:
+        """差值超阈值但不统计显著判定为 WARN。"""
+        verdict = _judge_task_type(
+            bench_correct=2,
+            bench_total=3,
+            gen_correct=8,
+            gen_total=20,
+            tolerance=0.10,
+            alpha=0.05,
+        )
+        assert verdict == "WARN"
+
+
+# ---------------------------------------------------------------------------
+# TestCalibrateIntegration
+# ---------------------------------------------------------------------------
+
+
+class TestCalibrateIntegration:
+    """calibrate 辅助函数集成测试。"""
+
+    def test_baseline_params_must_be_paired(self) -> None:
+        """baseline 参数必须成对出现。"""
+        with pytest.raises(ValueError, match="成对"):
+            _validate_calibrate_args(baseline_db="some.db", baseline_run_id=None)
+
+    def test_baseline_params_both_none_ok(self) -> None:
+        """两个参数都为 None 不报错。"""
+        _validate_calibrate_args(baseline_db=None, baseline_run_id=None)
+
+    def test_baseline_params_both_provided_ok(self) -> None:
+        """两个参数都提供不报错。"""
+        _validate_calibrate_args(baseline_db="some.db", baseline_run_id="run-001")
+
+    def test_baseline_run_id_only_raises(self) -> None:
+        """只提供 run_id 也报错。"""
+        with pytest.raises(ValueError, match="成对"):
+            _validate_calibrate_args(baseline_db=None, baseline_run_id="run-001")
+
+    def test_has_fail_returns_exit_code_1(self) -> None:
+        """存在 FAIL 时返回退出码 1。"""
+        verdicts = {"Object Recognition": "PASS", "Action Reasoning": "FAIL"}
+        assert _calibrate_exit_code(verdicts) == 1
+
+    def test_all_pass_or_warn_returns_exit_code_0(self) -> None:
+        """全部 PASS 或 WARN 时返回退出码 0。"""
+        verdicts = {"Object Recognition": "PASS", "Spatial Perception": "WARN"}
+        assert _calibrate_exit_code(verdicts) == 0
+
+    def test_all_pass_returns_exit_code_0(self) -> None:
+        """全部 PASS 时返回退出码 0。"""
+        verdicts = {"Object Recognition": "PASS", "Action Reasoning": "PASS"}
+        assert _calibrate_exit_code(verdicts) == 0
+
+    def test_empty_verdicts_returns_exit_code_0(self) -> None:
+        """空 verdicts 时返回退出码 0。"""
+        assert _calibrate_exit_code({}) == 0