---
id: 20260827-T0-03
title: "新方法FREF：用函数级执行反馈优化代码生成模型"
title_en: "FREF: Function-Level Execution Feedback Boosts Code LLMs"
url: https://ai.daily.yangsir.net/daily/20260827-T0-03
issue_date: 2026-08-27
publish_date: 2026-08-26T04:00:00.000Z
category: research
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2608.23632
---

# 新方法FREF：用函数级执行反馈优化代码生成模型

一篇新论文提出FREF（函数级执行反馈）方法，用于代码偏好优化。与过程监督在数学推理中的成功不同，代码生成一直缺乏标准化的步骤定义。FREF通过函数级执行反馈，为代码生成模型提供更细粒度的指导，有望提升生成代码的正确性。

## English Version

**FREF: Function-Level Execution Feedback Boosts Code LLMs**

A new paper introduces FREF (Function-Level Execution Feedback) for code preference optimization. Unlike process supervision in mathematical reasoning, code generation lacks standard step definitions. FREF provides more granular guidance via function-level execution feedback, potentially improving generated code correctness.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2608.23632)

**详情页**：https://ai.daily.yangsir.net/daily/20260827-T0-03

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*