EvalResearchBench (ERB) is a benchmark that tests whether AI agents can autonomously design evaluation tasks for recursive self‑improvement. In ERB, a researcher agent selects or synthesizes tasks, builds graders, and refines them through pilot tests before producing a final evaluator that runs within fixed time and API budgets. Experiments with nine researchers and thirteen candidate models show that the best automated evaluators achieve about 75% concordance with target benchmarks, below the 91% ceiling set by target disagreements, and that human‑designed task samples still outperform many automated approaches.
By Yaolun Zhang, Tianyi Xu, Yujie Zhao, Jishen Zhao, Qingyun Wu, Huazheng Wang
The paper introduces VERSE, a Verified Self‑Evolving optimizer that enhances LLM agent harnesses by allowing the optimizer to test edits, replay failures, and perturb steps while tracking fixes and regressions. VERSE builds its own tools for failure analysis, verification, training audits, and workflow control, and uses this feedback to revise the harness’s prompts, skills, tools, hooks, and notes without changing model weights. In experiments across five executors and multiple languages, VERSE improves all evaluated harness optimizers, achieving higher accuracy on held‑out and out‑of‑distribution tasks compared to the strongest baselines.
By Zekai Wang, Yingqiang Ge, Zekun Wang, Hai Wang, Yuhui Xu, Joshua Frandsen, Shancong Fu, Ashia C. Wilson, Chandan K. Reddy
arXiv:2610.14224v1 Announce Type: new
Abstract: LLM agents' capabilities depend not only on their underlying models but also on their execution harnesses. Recursive self-improvement (RSI) offers a pr...
By Ruiyao Xu, Yanfei Chen, Zhongying CuiZhu, Bhavana Dalvi Mishra, Yifei Ming, Han Yu, Rujun Han, Chen-Yu Lee, Tomas Pfister
arXiv:2608. 05519v1 Announce Type: new Abstract: Agent benchmarks usually measure task completion and treat resource use as an auxiliary statistic.
By Jie Wu, Ming Gong, Feixiang Cheng, Qinqin Zhao
The paper introduces Regularized Recursive Self-Improvement of Agent Harnesses (RRSI), a method that applies regularization principles to the iterative editing of an LLM agent’s harness—prompts, control flow, tooling, memory, and context management. RRSI limits the number of edits per candidate, encourages novel trajectories, and uses a critic and pruner to filter out benchmark‑specific or ineffective changes, thereby favoring reusable agent mechanisms. Experiments on eight benchmarks show RRSI improves performance by up to 14.1 points on the training split and 4.7 points on out‑of‑distribution tests, while reducing policy token usage by 30% compared to unregularized evolution.
By Peng Xia, Rujun Han, Zifeng Wang, Yanfei Chen, Yufan Zhang, Yoonho Lee, Chengsong Huang, Han Yu, Zhongying CuiZhu, Yifei Ming, Huaxiu Yao, Burak Gokturk, Tomas Pfister, Chen-Yu Lee
The paper introduces ExTS, a tree‑search policy designed for budget‑constrained agentic search where evaluation and generation costs are high. ExTS treats expansion as a value‑of‑information decision, combining discriminative reward shaping, a stochastic virtual child, and quality‑conditioned branching to allocate budget more effectively. Experiments on prompt optimization, code generation, molecular structure elucidation, and agentic workflow optimization show ExTS matching or surpassing task‑specific baselines with an average gain of +5.5% using a single configuration, and the authors also present pilot‑run diagnostics to guide adaptation to different problem structures.
By Haoyang Fang, Bernie Wang
RobustSGPO extends semantic‑gradient‑based prompt optimization (SGPO) by explicitly defining the edit scope, constructing and validating patches, and allowing search to resume from either the current or retained snapshots. In the AgentX brainstorming workflow, experiments on 120 tasks and 95 runs show that periodic permission scheduling improves test scores, while RobustSGPO raises completion rates on held‑out tasks from 60.0% to 80.0% and test quality from 3.77 to 4.14 within a 20‑million‑token budget. The approach also demonstrates that category‑based retention mitigates source‑task degradation after a shift, and that search‑space control enhances quality through executable edits and alternative starting points.
By Zibo Zhao, Jijun Shi, Mo Zhou, Zhongyuan Wang, Shifu Bie, Yunfei Zhang, Xuanting Zhou, Xiangyu Wu, Bin Liu, Ruiming Tang, Wenwu Ou, Kun Gai
arXiv:2609.33180v2 Announce Type: replace-cross
Abstract: As recursive self-improvement (RSI) rapidly advances, reliable evaluation becomes critical for guiding adaptive search. RSI typically relies...
By Xiaojing Sun, Yuhan Zeng, Zihua She, Xiao Wang
arXiv:2610.18813v1 Announce Type: new
Abstract: Harness self-evolution revises an agent's software scaffolding, including prompts, memory, workflows, and executable components, while keeping model we...
By Chuyang Wei, Haoxiang Guan, Minghang Zhu, Kexin Zhou, Jiyan He, Xiren Zhou, Huanhuan Chen, Yifei Shen
arXiv:2608. 09629v1 Announce Type: new Abstract: Self-evolving agents are usually built around prescribed optimization pipelines: the framework decides how to gather evidence, revise a persistent artifact, select candidates, and stop.
By Hui Xue, Fan Yang
Recent systems for optimizing prompts, programs, and ML workflows typically rely on explicit outer-loop controllers such as evolutionary search, bandits, or textual-gradient methods. We ask a fundamentally different question: how much of this search policy can be internalized by a single tool-using agent?
The paper argues that evaluating agents as fixed models is insufficient, proposing instead to treat them as configurable systems. Using a new benchmark of four scientific tasks, the authors analyze how five configuration aspects—task information, reasoning, self‑verification, time budget, and backbone model—affect performance, noting that about 54% of outcome variance arises from run‑to‑run differences even with the same settings. The study finds that providing more task information has the strongest impact, while interactions among settings (e.g., extra time only helps with adequate information or model capability) and the choice of verification tools significantly shape agent behavior.
By Luis Wiedmann, Leander Girrbach, Cordelia Schmid, Zeynep Akata