arXiv cs.AI· Minhua Lin, Hanqing Lu, Zhan Shi, Bing He, Rui Mao, Zhiwei Zhang, Zongyu Wu, Xianfeng Tang, Hui Liu, Zhenwei Dai, Rui Zhang, Xiang Zhang, Suhang Wang, Benoit Dumoulin, Jian Pei·· 11 小时前AI 评分31
立场文章：智能体演化是 LLM 持续进化的路径
Position: Agentic Evolution is the Path to Evolving LLMs
AI 导读
一篇立场文章提出，LLM 从固定训练集走向开放真实环境后，静态训练无法跟上部署环境变化，需要新增“演化”这一扩展轴。作者主张将演化从固定流程升级为自主演化智能体，并提出 A-Evolve 框架，把部署时改进视为对持久系统状态的目标导向优化。文章还提出演化扩展假设：适应能力随分配给演化的算力扩展，代码已公开。
正文 · 原文
Authors:Minhua Lin, Hanqing Lu, Zhan Shi, Bing He, Rui Mao, Zhiwei Zhang, Zongyu Wu, Xianfeng Tang, Hui Liu, Zhenwei Dai, Rui Zhang, Xiang Zhang, Suhang Wang, Benoit Dumoulin, Jian Pei
Abstract:As Large Language Models (LLMs) move from curated training sets into open-ended real-world environments, a fundamental limitation emerges: static training cannot keep pace with continual deployment environment change. Scaling training-time and inference-time compute improves static capability but does not close this train-deploy gap. We argue that addressing this limitation requires a new scaling axis-evolution. Existing deployment-time adaptation methods, whether parametric fine-tuning or heuristic memory accumulation, lack the strategic agency needed to diagnose failures and produce durable improvements. Our position is that agentic evolution represents the inevitable future of LLM adaptation, elevating evolution itself from a fixed pipeline to an autonomous evolver agent. We instantiate this vision in a general framework, A-Evolve, which treats deployment-time improvement as a deliberate, goal-directed optimization process over persistent system state. We further propose the evolution-scaling hypothesis: the capacity for adaptation scales with the compute allocated to evolution, positioning agentic evolution as a scalable path toward sustained, open-ended adaptation in the real world. Our code is publicly available at this https URL.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2602.00359 [cs.AI] |
| (or arXiv:2602.00359v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2602.00359 arXiv-issued DOI via DataCite |
Submission history
From: Minhua Lin [view email]
[v1]
Fri, 30 Jan 2026 22:15:58 UTC (12,158 KB)
[v2]
Fri, 13 Mar 2026 22:12:22 UTC (12,156 KB)
[v3]
Wed, 4 Nov 2026 21:53:37 UTC (12,170 KB)
来源：arXiv cs.AI · arxiv.org