ContextEvo, a framework that learns a context policy from long-horizon trajectories and reconstructs the model-visible context at key decision points, identifies context-related failures, and applies targeted policy updates, is introduced.
Abstract
Harness evolution improves LLM agents by learning from execution trajectories, but existing experience- and skill-based methods are less effective on long-horizon tasks. As interactions grow, useful evidence can be buried by redundant or outdated context, making context management itself a key bottleneck. We introduce ContextEvo, a framework that learns a context policy from long-horizon trajectories. ContextEvo reconstructs the model-visible context at key decision points, identifies context-related failures, and applies targeted policy updates. Starting from the open-source Pi-agent harness, ContextEvo improves performance across three long-horizon task benchmarks, achieving results comparable to or better than several prominent agent harnesses, including Codex, OpenCode, and OpenClaw. Additional analyses show that fixed or locally evolved context strategies can fall short under long-horizon information pressure, while our methods adapt to the information demands of each environment.
Experiments across diverse agent benchmarks show that \textit{Experience Funnel} consistently improves agent capability over state-only evolution and policy-internalization approaches, while progressively converting useful explicit experience into autonomous policy competence.
Wen-Bo Gao, Zhao-Mou Song, Zhi-Yuan Ji et al.· arXiv.org· 1 citation
LLM-based long-horizon agentic post-training is often bottlenecked by rollout generation: trajectories span many interaction turns, completion times vary substantially, and synchronous update barriers leave faster workers waiting for stragglers. Asynchronous reinforcement learning which has been adopted in LLM post-tra...
William Hoy, Jing-Xuan Fan, Nurcin Celik et al.· 0 citations
Interactive language-model agents increasingly solve complex tasks through long-horizon, multi-call reasoning, where errors in beliefs or actions can compound across tool interactions. Retrospective search can recover from such failures but is prone to misallocation. Delayed outcomes obscure the contribution of interme...
Xin-Yue Zeng, Shivam Shandilya, Guilherme Potje et al.· 0 citations
The concept of in-context self-evolution is formalized and VALVE, a validated-gated framework for long-horizon skill optimization is introduced, which establishes finite convergence, provides theoretical guarantees for future-task gain and drawdown, and derive the validation and evaluation holdout sizes required for a...
Yi-Fan Wang, Hao Cheng, Xiao-Min Li et al.· arXiv.org· 1 citation
This formulation enables a systematic study of key self-improvement factors through the proposed Evo-Harness, and provides a principled understanding of how LLM agents can effectively learn on the fly.
Tian-Xin Wei, Zhan Shi, Min-hua Lin et al.· 14 citations
PILOT is presented, a supervisor-worker harness for live self-improvement through two coupled mechanisms: (1) live steering lets a separate supervisor redirect or abort the active worker during execution; and (2) live self-evolution distils procedures and failure modes revealed during execution into reusable skills and...
Yang Xiao, Yu-Song Sun, Haoming Wu et al.· 2 citations