The paper introduces ANCHOR, an external supervisory framework driven by large language models (LLMs) that provides evaluative feedback at multiple stages of self‑evolving agents. By integrating ANCHOR into two open‑source self‑evolving agent frameworks, the authors demonstrate that it significantly improves safety performance while preserving core capabilities across coding, mathematical reasoning, and safety tasks. The study also finds that supervision based on execution results is especially effective and that increasing supervision frequency yields diminishing returns, offering practical guidance for future research.
By Dianxing Shi, Bowen Wang, Junqi He, Junhao Chen, Yuta Nakashima
SEABench is a benchmark designed to study endogenous misalignment in self‑evolving large language model agents. It contains 48 longitudinal task sequences across various evolution surfaces, task domains, and harm types, and includes an adaptive trajectory discovery pipeline that probes for failures while preserving task intent. Evaluations show that self‑evolution improves task completion rates but often introduces safety failures absent in non‑evolving baselines, with divergent safety behaviors reflected in agents’ chain‑of‑thought reasoning that can be monitored to mitigate unsafe actions.
The paper introduces the concept of Evolutionary Safety for recursive self-improving AI, focusing on how safety properties evolve as an AI system and its successors change. It identifies key risks such as intent drift, error accumulation, and safety-property erosion, and presents a taxonomy covering agent state, model state, evaluation, environment, and update mechanisms. The authors propose methods for discovering and evaluating evolutionary risks, and outline governance principles for modification, selection, authorization, provenance, and recovery, while highlighting open problems for maintaining safety in persistent, adaptive, and recursively self-improving systems.
By Chang Gong, Jingping Bi, Di Yao, Xinjian Liang, Chao Xiang, Ruijie Guo
arXiv:2609.35596v2 Announce Type: replace-cross
Abstract: Self-evolving LLM agents have gained prominence for their ability to improve after deployment by modifying their harness, including their con...
By Saswat Das, Parvati Viswanathan, Daniel Donnelly, Chang Huang, Sahar Abdelnabi, Ferdinando Fioretto
arXiv:2606. 28347v1 Announce Type: cross Abstract: Contemporary AI safety spans pre-training interventions, post-training alignment, deployment-time controls, monitoring, and red-teaming.
By Charles L. Wang, Keir Dorchen, Peter Jin
arXiv:2608. 09885v1 Announce Type: new Abstract: The safety of large language model (LLM) agents depends not only on model weights but also on the agent harness that manages context, memory, tools, permissions, and runtime control.
By Wanying Qu, Qinghua Mao, Yu Li, Jiyao Liu, Xin Zhang, Dadi Guo, Yanxu Zhu, Qingyu Liu, Leitao Yuan, Xi Lin, Shanfeng Zhu, Yanwei Fu, Jing Shao, Xia Hu, Dongrui Liu
arXiv:2607. 13104v1 Announce Type: new Abstract: Self-improving autonomous agents are moving from research prototypes to deployed systems.
By Zhe Ren, Yimeng Chen, Dandan Guo, Guowei Rong, Tonghui Li, R. B. Xiong, Qingfeng Lan, Wenyi Wang, Li Nanbo, Yibo Yang, Mingchen Zhuge, J\"urgen Schmidhuber
arXiv:2607. 28691v1 Announce Type: cross Abstract: Personalized AI agents are often configurable without giving users control over the artifacts that determine their future behavior.
By Roy Zhao (Paul G. Allen School of Computer Science & Engineering, University of Washington), Zhenyu Zhao (Independent Researcher)
arXiv:2608. 13120v1 Announce Type: new Abstract: Agent Skills are today either hand-authored or produced in a single LLM generation pass, and consequently possess no closed loop through which they might improve from the interaction failures they actually cause.
By Qianxi Yan, Chunrong Chen, Jiuzhou Zhao, Min Zhang, Yongzhou Xu, Xiaochuan Xu
SafeEvolve is an experience-driven framework that co‑evolves a harness and policy to improve safety alignment for LLM‑based agents. It uses on‑policy trajectory safety evidence to update safety prompts and hierarchical skills, producing auditable harness artifacts. The policy is trained via a two‑stage SFT‑RL pipeline that bootstraps with the evolved harness and then refines behavior through verifier‑decomposed rewards, yielding a better safety‑utility tradeoff on benchmarks such as AgentDojo.
HarnessEvolve is a self‑evolving framework that improves agent harnesses—prompts, skills, tools, and execution logic—by learning from reference trajectories. It separates execution, evaluation, optimization, and gating into independent modules, addressing credit assignment failure, shortcut learning, and catastrophic forgetting. The approach uses reference trajectories to extract error signals, applies quality and performance gates to candidate updates, and validates updates on held‑out data, consistently outperforming state‑of‑the‑art baselines across diverse benchmarks.
By Wen Jiang, Mingmin Chu, Yimeng Tian, Qianxin Zhang, Haofei Yang, Rui Yang, Yang Liu, Tao Lv, Fangming Li
The article surveys safety concerns for self‑evolving agents that continually update their internal state, such as model parameters and memories, from new interactions. It introduces the SAVER framework, which tracks reusable influence, adaptation, violations, exposure, and response to assess whether safety properties persist as agents evolve. The survey finds that legitimate state can become unsafe when its persistence, authority, or scope expands beyond its original conditions, and highlights gaps in current research on descendant repair and longitudinal evaluation.
By Jiahao Chen, Zhou Feng, Oubo Ma, Yichen Yan, Ruixiao Lin, Hangtao Zhang, Linkang Du, Hengyu An, Yong Yang, Jun Liu, Junhao Li, Naen Xu, Chunyi Zhou, Yuan Su, Zehao Jin, Qianli Ma, Leyi Qi, Yiming Wang, Zhe Ma, Yuwen Pu, Mengyao Du, Yuanyi Song, Enhao Huang, Zhihui Fu, Jun Wang, Jinfeng Li, Yuefeng Chen, Hui Xue, Yiming Li, Tianyu Du, Shouling Ji