arXiv：cs.CL· Shangyin Tan, Ben Zhou, Lakshya A Agrawal, Jialin Zhang, Yilin Chen, Qiuyang Mang, Donghyun Lee, Dennis Jacob, Koushik Sen, Alex Dimakis, Matei Zaharia·· 6 小时前
让 LLM 智能体学会遗忘：突破自动化研究瓶颈
Scaling Discovery Requires Forgetting
AI 导读
研究发现，LLM 智能体无论延长迭代、并行运行还是互相通信，都会因历史积累而锚定于自身过往尝试，收益趋于停滞。为此作者提出定期丢弃搜索状态（包括各智能体会话与候选历史），从当前最优候选重启，在启发式编程、内核优化和模型训练四类任务上比并行智能体提升最多 15%，比单智能体提升最多 39%。
正文
Authors:Shangyin Tan, Ben Zhou, Lakshya A Agrawal, Jialin Zhang, Yilin Chen, Qiuyang Mang, Donghyun Lee, Dennis Jacob, Koushik Sen, Alex Dimakis, Matei Zaharia
Abstract:Automated research with LLM agents has shown promising results in various domains. Through iterative refinement and scored verifiers, LLM agents are capable of discovering new ideas. There are different ways to scale automated research: let one agent iterate for longer, run several agents in parallel, or further let the parallel agents communicate and build on one another's candidates. However, we find that all three strategies eventually plateau as the agents accumulate history. Each agent becomes anchored to its own past attempts, making fewer and smaller edits. Beyond this point, additional budget brings little further improvement. In this paper, we show that the key to scaling discovery further is to forget. Periodically discarding the search state, including every agent's session and the candidate history, and restarting from the best candidate found so far, unblocks the search and pushes the same budget further.
We implement forgetting in a team of agents with communication and evaluate it on four tasks spanning heuristic programming, kernel optimization and model training. The results show up to 15% improvement compared to parallel agents, and up to 39% compared to a single agent. Our results suggest that, for open-ended discovery, what an agent is allowed to remember matters as much as how much compute it is given.
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.20563 [cs.AI] |
| (or arXiv:2610.20563v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.20563 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Shangyin Tan [view email]
[v1]
Wed, 14 Oct 2026 17:49:37 UTC (114 KB)
来源：arXiv：cs.CL · arxiv.org