English(EN) The Tests Passed. The Function Already Existed.

新的RepoExec基准测试揭示LLM在代码依赖准确性方面存在困难

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-15 22:12

来自FPT Software AI Center的一篇论文介绍了一个名为RepoExec的新基准测试，该测试通过考虑正确性和依赖调用率来评估大型语言模型在代码生成方面的能力。基准测试显示，当前模型在准确利用提供的依赖项方面存在困难，许多模型未能调用正确的依赖项或重写现有函数而不是解决问题。研究还发现，仅提供函数签名和文档字符串而不提供函数体，可能导致模型将任务误解为少样本示例，从而生成空代码或错误代码。 AI

影响突出了LLM代码生成能力的一个关键差距，表明当前模型可能无法可靠地集成外部代码依赖项。

排序理由介绍LLM代码生成新基准的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 dev.to — LLM tag 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 27 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 介绍LLM代码生成新基准的研究论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, product; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

dev.to — LLM tag TIER_1 English(EN) · hidetzu · 2026-09-15 22:12

测试已通过。该功能早已存在。

<p>Both models passed the tests. Both rewrote a function that was sitting in their own prompt, twenty lines up.</p> <p>That is a worked example from the RepoExec paper, and it is the reason the benchmark exists. <code>pass@k</code> cannot see it — the tests are green. The bill ar…

本文如何被排名

报道来源 [1]

测试已通过。该功能早已存在。

相关实体

相关话题