Tracing the Thoughts of a Coding Agent Playing ARC-AGI-3: Lessons for Continual Learning
作者: Chen Wu, Josh Passenger, Yin Song
分类: cs.AI
发布日期: 2026-10-08
备注: Accepted at the NeurIPS 2026 Workshop on Continual Learning in the Era of Foundation Models and Embodied Agents (CL4FMAgents)
💡 一句话要点
提出一种测量协议以追踪编码代理在ARC-AGI-3中的学习过程
🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 编码代理 连续学习 抽象推理 知识管理 思维追踪 自动化编程 智能系统
📋 核心要点
- 现有方法在处理连续学习时，往往无法有效管理知识的遗忘与保留，导致性能下降。
- 本论文提出了一种测量协议，通过追踪编码代理在不同任务中的思维过程，分析其学习机制。
- 实验结果表明，代理在新任务中几乎不重用旧脚本，74%的脚本在任务边界前被放弃，显示出其知识重写的能力。
📝 摘要（中文）
本研究探讨了一种编码代理在一系列抽象推理任务中如何学习。该代理在固定的环境中运行，通过编写和执行Python及Shell脚本进行操作。代理在每个回合之间不保留状态，仅保留其编写的文档，因此每个形成、修正或放弃的思维都会留下痕迹。我们让代理参与ARC-AGI-3，这是一组没有说明的互动推理游戏。每个游戏由多个关卡组成，清除一个关卡的策略可能在下一个关卡失效，因此每个新关卡实际上都是一个新任务。我们的贡献在于提出了一种测量协议，能够追踪每个思维在这些文件中的形成、修正或放弃过程，并在七次评估运行中应用于来自两个模型家族的三种基础模型。
🔬 方法详解
问题定义：本论文旨在解决编码代理在连续学习中如何有效管理知识的遗忘与保留的问题。现有方法往往无法追踪思维过程，导致知识的混乱与错误。
核心思路：论文的核心思路是通过记录代理在不同任务中的思维过程，分析其如何形成、修正或放弃知识，从而实现有效的知识管理。这样的设计使得研究者能够清晰地了解代理的学习轨迹。
技术框架：整体架构包括代理的执行环境、任务记录系统和思维追踪协议。代理在固定环境中运行，通过编写脚本进行操作，所有的操作和观察都被完整记录。
关键创新：最重要的技术创新点在于提出了一种能够追踪编码代理思维过程的测量协议，这与现有方法的黑箱分析形成鲜明对比，提供了透明的学习过程分析。
关键设计：在设计中，代理的脚本几乎不跨任务重用，只有33个脚本在任务边界间被引用。代理重写知识时保留一般规则，丢弃特定关卡的细节，且笔记从不修改，形成了一个不断积累的知识库。
🖼️ 关键图片
📊 实验亮点
实验结果显示，代理在新任务中几乎不重用旧脚本，74%的脚本在任务边界前被放弃，且只有33个脚本跨越任务边界，表明其在知识重写和选择性遗忘方面的有效性。这种机制使得代理能够在复杂任务中保持较高的学习效率。
🎯 应用场景
该研究的潜在应用领域包括智能编程助手、自动化测试生成和教育领域的个性化学习系统。通过理解编码代理的学习过程，可以为未来的人工智能系统设计提供重要的理论基础和实践指导。
📄 摘要（原文）
We study how a coding agent learns across a sequence of abstract reasoning tasks. The agent runs on a frozen foundation model inside a fixed harness and acts by writing and running Python and shell scripts. It retains no state across turns other than its written artifacts, so every thought it forms, carries, corrects or abandons leaves a trace, where a thought is any belief, rule or plan committed to a file. We let the agent play ARC-AGI-3, a set of interactive reasoning games that provide no instructions. Each game is a sequence of levels, and a strategy that clears one level can fail on the next, so every new level is in effect a new task. The agent records what it learns as Python scripts and text notes, while the harness keeps a complete log of every action and observation. Our contribution is a measurement protocol that traces each thought through these files, from the task where it forms to the task where it is corrected or abandoned, applied to seven evaluation runs with three backbones from two model families. Scripts written for one task are almost never called again in a later task (33 of 630 references cross a task boundary), because most scripts embed the state of the current level. Instead, the agent rewrites its knowledge into new scripts, keeping the general rules and dropping the level-specific details, and abandons 74% of the scripts it wrote before a boundary. The notes, which only the model reads, are never revised: the agent appends without removing earlier claims, and the contradictions that accumulate are settled against the log. Because the log preserves everything, the agent forgets selectively, not catastrophically. The most costly error is a hard-coded value carried into a task where it no longer holds. These findings come from the files the agent wrote, without access to the model, and constitute a white-box analysis of how a coding agent continually learns.