arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2601.04620cs.AI
收藏

AgentDevel: 将自演化大语言模型代理重新定义为发布工程

AgentDevel: Reframing Self-Evolving LLM Agents as Release Engineering

Fudan University(复旦大学)

Di Zhang

AI总结:

AgentDevel将LLM代理改进重新定义为发布工程,通过无实现偏见的批评者、可执行诊断和翻转中心门控,实现稳定改进和可审计的发布流程。

摘要:

近年来,大语言模型(LLM)代理的进步主要集中在将自我改进机制嵌入代理或在许多并发变体中进行搜索。尽管这些方法可以提高整体分数,但往往会产生不稳定且难以审计的改进轨迹,使得难以保证不退化或在不同版本之间推理失败。我们重新定义代理改进为"发布工程":将代理视为可发布的制品,并将改进外部化为一个具有回归意识的发布流程。我们引入了AgentDevel,一个发布工程流程,该流程迭代运行当前代理,从执行轨迹中生成无实现偏见、症状级别的质量信号,通过可执行诊断合成一个单一的发布候选版本(RC),并在翻转中心门控下推广。AgentDevel有三个核心设计:(i)一个无实现偏见的LLM批评者,可以表征失败出现而无需访问代理内部;(ii)基于脚本的可执行诊断,汇总主导症状模式并生成可审计的工程规范;(iii)翻转中心门控,优先考虑通过到失败的回归和失败到通过的修复作为一级证据。与基于种群的搜索或在代理内的自我完善不同,AgentDevel维护一条单一的规范版本线,并将不退化作为主要目标。在执行密集型基准测试中,实验表明AgentDevel在产生可重复、可审计的制品的同时,显著减少了回归,实现了稳定的改进。总体而言,AgentDevel为构建、调试和发布LLM代理作为软件开发提供了实用的开发纪律。

英文摘要:

Recent progress in large language model (LLM) agents has largely focused on embedding self-improvement mechanisms inside the agent or searching over many concurrent variants. While these approaches can raise aggregate scores, they often yield unstable and hard-to-audit improvement trajectories, making it difficult to guarantee non-regression or to reason about failures across versions. We reframe agent improvement as \textbf{release engineering}: agents are treated as shippable artifacts, and improvement is externalized into a regression-aware release pipeline. We introduce \textbf{AgentDevel}, a release engineering pipeline that iteratively runs the current agent, produces implementation-blind, symptom-level quality signals from execution traces, synthesizes a single release candidate (RC) via executable diagnosis, and promotes it under flip-centered gating. AgentDevel features three core designs: (i) an implementation-blind LLM critic that characterizes failure appearances without accessing agent internals, (ii) script-based executable diagnosis that aggregates dominant symptom patterns and produces auditable engineering specifications, and (iii) flip-centered gating that prioritizes pass to fail regressions and fail to pass fixes as first-class evidence. Unlike population-based search or in-agent self-refinement, AgentDevel maintains a single canonical version line and emphasizes non-regression as a primary objective. Experiments on execution-heavy benchmarks demonstrate that AgentDevel yields stable improvements with significantly fewer regressions while producing reproducible, auditable artifacts. Overall, AgentDevel provides a practical development discipline for building, debugging, and releasing LLM agents as software development.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑