arXiv：cs.AI· Chen Wu, Josh Passenger, Yin Song·· 3 小时前
编程智能体玩 ARC-AGI-3 的思维轨迹追踪：持续学习启示
Tracing the Thoughts of a Coding Agent Playing ARC-AGI-3: Lessons for Continual Learning
AI 导读
研究追踪编程智能体在 ARC-AGI-3 交互推理游戏中的持续学习过程，该智能体基于冻结的基础模型运行在固定 harness 中，仅通过编写 Python 和 shell 脚本行动，除书面产物外不保留任何跨轮状态。
正文
Abstract: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.
| Comments: | Accepted at the NeurIPS 2026 Workshop on Continual Learning in the Era of Foundation Models and Embodied Agents (CL4FMAgents) |
| Subjects: | Artificial Intelligence (cs.AI) |
| ACM classes: | I.2.6; I.2.8; I.2.2 |
| Cite as: | arXiv:2610.11450 [cs.AI] |
| (or arXiv:2610.11450v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.11450 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Chen Wu [view email]
[v1]
Thu, 8 Oct 2026 08:04:18 UTC (64 KB)
来源：arXiv：cs.AI · arxiv.org