From 228c8e750ce410d2cb0ac933a7859a04083657f3 Mon Sep 17 00:00:00 2001 From: Sinonkey Date: Thu, 27 Aug 2026 19:39:54 +0800 Subject: [PATCH] =?UTF-8?q?docs(score=5Fdiff):=20=E8=B0=83=E6=95=B4?= =?UTF-8?q?=E8=AE=BE=E8=AE=A1=E7=BA=A6=E5=AE=9A=E8=A1=A8=E8=BF=B0=E4=B8=8E?= =?UTF-8?q?=20import=20=E6=A0=BC=E5=BC=8F?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- dm_agent/benchmarks/score_diff.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/dm_agent/benchmarks/score_diff.py b/dm_agent/benchmarks/score_diff.py index eacca75..062cabb 100644 --- a/dm_agent/benchmarks/score_diff.py +++ b/dm_agent/benchmarks/score_diff.py @@ -3,7 +3,7 @@ 这是「改了策略之后到底有没有变好」的判据。只读已有的 JSON 报告， 不跑 benchmark、不调模型、不联网——和 ``manifest_diff`` 同样是纯离线工具。 -三条设计约定： +三条设计约定如下： - **逐题翻转比总分更有指导性**。总分从 0.46 变 0.54 只说明「多过了一题」， 而「哪题修好了、哪题跑坏了」才指向下一步该看什么。回归（原本过、现在挂） @@ -16,7 +16,7 @@ from __future__ import annotations import argparse -import json +import json import sys from dataclasses import dataclass, field from pathlib import Path