PulseAugur
中
实时 17:11:16
English(EN) Where an agent's tokens go: 9 real GitHub bugs, two Gemini Flash-Lite models, 1 fix each

Gemini Flash-Lite 模型在智能体循环测试中各修复一个 GitHub Bug

一项实验测试了两个 Gemini Flash-Lite 模型修复真实 GitHub Bug 的能力,每个模型在九次尝试中成功解决了一个 Bug。Gemini 3.5 Flash-Lite 模型进行了 964 次调用,需要 17 次尝试;而 Gemini 3.1 Flash-Lite 模型调用次数较少,但重试次数更多,仅成功一次。每个修复 Bug 的成本估计在 0.43 美元至 0.56 美元之间,其中 token 使用量和缓存对总体费用影响显著,尤其是在智能体循环中重新发送上下文时。 AI

影响 为 LLM 智能体用于自动化软件 Bug 修复的成本和有效性提供了见解。

排序理由 该条目详细介绍了测试 LLM 在特定任务上能力的一项实验,包括成本分析和性能指标,这与研究相符。[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 →

Gemini Flash-Lite 模型在智能体循环测试中各修复一个 GitHub Bug

本文如何被排名

Signal score
48 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
该条目详细介绍了测试 LLM 在特定任务上能力的一项实验,包括成本分析和性能指标,这与研究相符。[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
product, other
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]

  1. dev.to — LLM tag TIER_1 English(EN) · AICostbudget.com ·

    代理的token去哪儿了:9个真实的GitHub Bug,两个Gemini Flash-Lite模型,各修复一个

    <p>I gave two Gemini Flash-Lite models the same 9 real bug reports from GitHub and let each one work in a loop: search, read a file, edit, run the tests, submit. Each model fixed one bug. Between them they made 964 model calls.</p> <p>The tasks come from SWE-bench-Live (github.co…