From a0cdb166176f3d2c85ad17244b4e724984c9dabd Mon Sep 17 00:00:00 2001
From: ThearchyHelios <work@thearchyhelios.com>
Date: Fri, 16 Oct 2026 17:11:09 +0800
Subject: [PATCH] =?UTF-8?q?=E9=87=8D=E8=AF=84=E6=97=B6=E6=8A=8A=E8=AF=84?=
 =?UTF-8?q?=E5=88=86=E8=A7=84=E5=88=99=E7=89=88=E6=9C=AC=E8=AE=B0=E8=BF=9B?=
 =?UTF-8?q?=E6=8C=87=E7=BA=B9=EF=BC=9B=E5=8E=BB=E6=8E=89=E5=87=A0=E5=A4=84?=
 =?UTF-8?q?=E6=B2=A1=E7=94=A8=E7=9A=84=E5=AF=BC=E5=85=A5?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

为什么：按新规则重评之后，和别的评测比时「评分规则」那一项要照实标出变了；总表上另记重评的时间。
---
 backend/app/eval/rescore.py         | 7 +++++--
 backend/app/eval/runner.py          | 5 +----
 backend/tests/test_eval_schedule.py | 4 ++--
 backend/tests/test_eval_score.py    | 6 ++++--
 backend/tests/test_eval_stats.py    | 2 --
 5 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/backend/app/eval/rescore.py b/backend/app/eval/rescore.py
index f1fda1f3..355609da 100644
--- a/backend/app/eval/rescore.py
+++ b/backend/app/eval/rescore.py
@@ -13,7 +13,7 @@ from typing import Any
 
 from sqlalchemy import select
 
-from app.db.base import SessionLocal
+from app.db.base import SessionLocal, utcnow
 from app.db.models_eval import EvalCase, EvalResult, EvalRun
 
 logger = logging.getLogger(__name__)
@@ -66,7 +66,10 @@ async def rescore(eval_run_id: str) -> int:
                     setattr(row, key, value)
         await session.flush()
         run = await session.get(EvalRun, eval_run_id)
-        await runner.refresh_summary(session, run)
+        # 指纹里的评分规则版本跟着换：和别的评测比时照实标出「评分规则」一项变没变
+        run.fingerprint = {**dict(run.fingerprint or {}), "scorer_version": score.SCORER_VERSION}
+        summary = await runner.refresh_summary(session, run)
+        run.summary = {**summary, "rescored_at": utcnow().isoformat()}
         await session.commit()
     return changed
 
diff --git a/backend/app/eval/runner.py b/backend/app/eval/runner.py
index c4c89be5..ca05f81b 100644
--- a/backend/app/eval/runner.py
+++ b/backend/app/eval/runner.py
@@ -33,9 +33,8 @@ from __future__ import annotations
 import asyncio
 import copy
 import logging
-import time
 from collections.abc import Mapping, Sequence
-from dataclasses import dataclass, field
+from dataclasses import dataclass
 from datetime import datetime
 from typing import Any
 
@@ -236,8 +235,6 @@ class _Arm:
     view: dict[str, Any] | None = None
     #: 这一臂的名字（A/B 时）
     name: str | None = None
-    #: 这一臂各项：{(题目 id, 第几次): 结果 id}
-    items: dict[tuple[str, int], str] = field(default_factory=dict)
 
 
 @dataclass
diff --git a/backend/tests/test_eval_schedule.py b/backend/tests/test_eval_schedule.py
index cab8b309..d3ad8bbb 100644
--- a/backend/tests/test_eval_schedule.py
+++ b/backend/tests/test_eval_schedule.py
@@ -13,7 +13,7 @@ from __future__ import annotations
 import asyncio
 import json
 import uuid
-from datetime import datetime, timedelta, timezone
+from datetime import timedelta, timezone
 
 import httpx
 import pytest
@@ -22,7 +22,7 @@ from sqlalchemy import delete, select
 
 from app.core.crypto import encrypt
 from app.db.base import SessionLocal, utcnow
-from app.db.models import NotifyChannel, Schedule, ScheduleFire
+from app.db.models import NotifyChannel, Schedule
 from app.db.models_eval import EvalResult, EvalRun
 from app.engine.runner import run_manager
 from app.eval import answer as answer_mod
diff --git a/backend/tests/test_eval_score.py b/backend/tests/test_eval_score.py
index 645f92b4..b9578a20 100644
--- a/backend/tests/test_eval_score.py
+++ b/backend/tests/test_eval_score.py
@@ -198,7 +198,7 @@ async def test_reading_the_sealed_answer_and_the_per_case_sql_match(engine_on, m
 async def test_rescoring_reads_the_seal_and_never_calls_a_model(engine_on, monkeypatch, tmp_path):
     from app.db.base import SessionLocal
     from app.db.models_eval import EvalResult, EvalRun
-    from app.eval import rescore
+    from app.eval import rescore, score
     from app.providers import mock_model
     from tests import eval_samples as es
     from tests import ledger_samples as ls
@@ -229,5 +229,7 @@ async def test_rescoring_reads_the_seal_and_never_calls_a_model(engine_on, monke
     async with SessionLocal() as session:
         row = (await session.execute(select(EvalResult).where(EvalResult.eval_run_id == er_id))).scalar_one()
         assert row.verdict == "pass" and row.scores[0]["outcome"] == "traced"
-        assert (await session.get(EvalRun, er_id)).summary["accuracy"]["passed"] == 1
+        rescored = await session.get(EvalRun, er_id)
+        assert rescored.summary["accuracy"]["passed"] == 1 and rescored.summary["rescored_at"]
+        assert rescored.fingerprint["scorer_version"] == score.SCORER_VERSION
     await es.drop_suites()
diff --git a/backend/tests/test_eval_stats.py b/backend/tests/test_eval_stats.py
index 7b4ecfe2..1b809646 100644
--- a/backend/tests/test_eval_stats.py
+++ b/backend/tests/test_eval_stats.py
@@ -7,8 +7,6 @@
 """
 from __future__ import annotations
 
-from datetime import timedelta
-
 import pytest
 
 import app.db.models  # noqa: F401 - conftest 按已导入的模型建表