Do Self-Evolving Agents Forget? Capability Degradation and Preservation in Lifelong LLM Agent Adaptation

• URL: http://arxiv.org/abs/2605.09315v1
• Date: Sun, 10 May 2026 04:20:24 GMT
• Title: Do Self-Evolving Agents Forget? Capability Degradation and Preservation in Lifelong LLM Agent Adaptation
• Abstract summary: We show that adapting to new task distributions can progressively degrade previously acquired capabilities across all major evolution channels.<n>We propose emphCapability-Preserving Evolution (CPE), a general stabilization principle that constrains destructive capability drift during continual adaptation.<n>Our findings suggest that stable long-horizon self-evolving agents require not only acquiring new capabilities, but also explicitly preserving previously learned ones.
• Score: 29.927620139561327
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent advances in LLM agents enable systems that autonomously refine workflows, accumulate reusable skills, self-train their underlying models, and maintain persistent memory. However, we show that such self-evolution is often non-monotonic: adapting to new task distributions can progressively degrade previously acquired capabilities across all major evolution channels. We identify this phenomenon as \emph{capability erosion under self-evolution} and show that it consistently emerges across workflow, skill, model, and memory evolution. To mitigate this issue, we propose \emph{Capability-Preserving Evolution} (CPE), a general stabilization principle that constrains destructive capability drift during continual adaptation. Across all four evolution dimensions, CPE consistently improves retained capability stability while preserving adaptation performance. For example, in workflow evolution, CPE improves retained simple-task performance from 41.8\% to 52.8\% under GPT-5.1 optimization while simultaneously achieving stronger complex-task adaptation. Our findings suggest that stable long-horizon self-evolving agents require not only acquiring new capabilities, but also explicitly preserving previously learned ones during continual adaptation.

Related papers

• EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness? [83.21711423118768]
  EVOHARNESSBENCH is a benchmark for evaluating agents under controlled harness evolution.<n>It contains 17 multi-stage harness streams constructed deterministically from verifier-based benchmarks.
  arXiv  Detail & Related papers  (2026-09-03T03:24:02Z)
• DIVE: Unlocking Self-Improvement in Frozen Language Models Through Diversity-Driven Skill Evolution [30.082997083095847]
  Large language models (LLMs) cannot retain post-deployment experience without parameter updates.<n>We introduce DIVE, a diversity-driven framework that enables frozen LLMs to improve by evolving from task experience and verifier feedback.
  arXiv  Detail & Related papers  (2026-08-12T18:06:41Z)
• Self-evolving LLM agents with in-distribution Optimization [60.05867547965365]
  Large Language Models (LLMs) have recently emerged as powerful controllers for interactive agents in complex environments.<n>We propose Q-Evolve, a self-evolving framework for LLM agents that unifies automatic process-reward labeling and policy learning.<n>We evaluate our method on AlfWorld, WebShop, and ScienceWorld, showing Q-Evolve outperforms strong baselines in sample efficiency, robustness, and overall task performance.
  arXiv  Detail & Related papers  (2026-06-05T15:09:52Z)
• MobEvolve: An Agentic Self-Evolving Heuristic System for Interpretable Human Mobility Generation [51.824145170051516]
  MobEvolve is a self-evolving framework for human mobility generation.<n>It employs an agent to iteratively evolve its internal logic.<n>It significantly outperforms state-of-the-art deep generative and LLM-based methods.
  arXiv  Detail & Related papers  (2026-06-01T03:46:25Z)
• CyberEvolver: Structured Self-Evolution for Cybersecurity Agents On the Fly [15.168064055560293]
  We introduce textscCyberEvolver, a self-evolving cybersecurity agent framework.<n>textscCyberEvolver iteratively revises its own scaffold based on experience from failed execution attempts.<n>We evaluate textscCyberEvolver on CTF challenges, vulnerability exploitation, and penetration-testing tasks using four open-source LLMs.
  arXiv  Detail & Related papers  (2026-05-25T16:26:59Z)
• Learning Without Losing Identity: Capability Evolution for Embodied Agents [14.412476605788482]
  We propose a capability-centric evolution paradigm for embodied agents.<n>We argue that a robot should maintain a persistent agent as its cognitive identity.<n>We present a unified framework in which capability evolution is decoupled from agent identity.
  arXiv  Detail & Related papers  (2026-04-09T04:51:07Z)
• SOLAR: A Self-Optimizing Open-Ended Autonomous Agent for Lifelong Learning and Continual Adaptation [3.239706565815636]
  Self-Optimizing Lifelong Autonomous Reasoner (SOLAR) is an open-ended autonomous agent.<n>It uses parameter-level meta-learning to self-improve, treating model weights as an environment for exploration.<n> Experiments demonstrate that SOLAR outperforms strong baselines on common-sense, mathematical, medical, coding, social and logical reasoning tasks.
  arXiv  Detail & Related papers  (2026-03-23T07:18:02Z)
• Position: Agentic Evolution is the Path to Evolving LLMs [56.733933092220845]
  We argue that addressing this limitation requires a new scaling axis-evolution.<n>Existing deployment-time adaptation methods lack the strategic agency needed to diagnose failures and produce durable improvements.
  arXiv  Detail & Related papers  (2026-01-30T22:15:58Z)
• Evolving LLMs' Self-Refinement Capability via Synergistic Training-Inference Optimization [53.93621974137829]
  Self-Refinement refers to a model's ability to revise its own responses to produce improved outputs.<n>EVOLVE is a framework for eliciting and tracking the evolution of Self-Refinement through iterative training.<n>We demonstrate the potential of leveraging Self-Refinement to achieve broader Self-Improvement of intrinsic model abilities.
  arXiv  Detail & Related papers  (2025-02-08T15:21:55Z)
• Agent Alignment in Evolving Social Norms [65.45423591744434]
  We propose an evolutionary framework for agent evolution and alignment, named EvolutionaryAgent. In an environment where social norms continuously evolve, agents better adapted to the current social norms will have a higher probability of survival and proliferation. We show that EvolutionaryAgent can align progressively better with the evolving social norms while maintaining its proficiency in general tasks.
  arXiv  Detail & Related papers  (2024-01-09T15:44:44Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.