arXiv cs.AI· Qile Jiang, George Em Karniadakis·· 12 小时前AI 评分58
SwarmEvolve：在隐空间通信的智能体群体用于进化式发现与优化
SwarmEvolve: Latent Agent Swarms for Evolutionary Discovery and Optimization
AI 导读
SwarmEvolve 提出让每轮数百个低成本智能体组成群体，在隐空间中推理而不解码为文本，并通过传递 KV cache 共享信息。在 21 个数学发现与科学计算问题上，它的平均得分最高，达到其他 SOTA 框架 60 分钟才能取得的最佳分数只需中位数 3 到 8 分钟；在部分问题上，使用 Qwen3 8B 的 SwarmEvolve 改进了 AlphaEvolve 最初用前沿 LLM 找到的解。
正文 · 原文
Abstract:LLM-guided evolutionary search uses AI agents to improve programs for mathematical and computational problems by iteratively refining solutions and combining high-scoring methods. However, the process is often slow because the LLM generates a long textual chain of thought for each round and must repeat this generation many times to maintain population diversity. We present SwarmEvolve, a framework in which hundreds of low-cost agents per round work in swarms, reason in latent space without decoding to text, and share information by passing KV caches. Each swarm studies the problem from a different angle. A novelty filter keeps the most distinct agent ideas by directly measuring the distances between the agents' hidden state vectors. Crossover of ideas happens naturally in latent space, as a summarizer agent receives the combined KV caches of all swarm members. On 21 problems in mathematical discovery and scientific computing, SwarmEvolve gets the highest average score and needs a median of only 3 to 8 minutes to reach the best scores that other state-of-the-art frameworks reach in 60 minutes. On several problems, SwarmEvolve with Qwen3 8B improves on the solutions that AlphaEvolve first found with frontier LLMs. Our work shows that agent swarms that communicate in latent space offer a fast, low-cost approach to program search for scientific problems.
| Subjects: | Artificial Intelligence (cs.AI); Neural and Evolutionary Computing (cs.NE) |
| Cite as: | arXiv:2611.08483 [cs.AI] |
| (or arXiv:2611.08483v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2611.08483 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Qile Jiang [view email]
[v1]
Thu, 5 Nov 2026 18:59:29 UTC (1,952 KB)
来源：arXiv cs.AI · arxiv.org