The paper introduces a co‑evolving framework where a target agent improves by learning from its own failures, and a separate failure agent is trained to generate hard negative trajectories. These hard negatives, derived from plausible but incorrect attempts, help the target agent better distinguish successful behavior from subtle errors. Experiments on online shopping, scientific reasoning, and interactive SQL querying show a 5.7% average reward improvement over baseline methods.
arXiv:2601.12310v2 Announce Type: replace Abstract: Self-training systems often degenerate due to the lack of an external criterion for judging data quality, leading to reward hacking and semantic dr...
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.
arXiv:2606. 07412v1 Announce Type: cross Abstract: LLM-driven software engineering agents have become a central testbed for real-world language-model capability, yet their training remains limited by the availability of high-quality SWE tasks.
arXiv:2608. 05628v1 Announce Type: new Abstract: Although agent skills equip LLMs with reusable procedural knowledge, manual maintenance suffers from high costs, unscalability, and misalignment.
arXiv:2607. 20668v1 Announce Type: cross Abstract: TextGrad improves language-model systems by revising text from feedback.
arXiv:2509. 03206v2 Announce Type: replace-cross Abstract: Learning from reward functions and imitation learning of demonstrations are the two principal approaches for training autonomous systems that interact with an environment through action and observation.
CAFE (Coupled Agent–Feedback Evolution) is a framework that lets a shared‑parameter model alternate between acting as a search agent and as a critic that provides corrective feedback. By learning when to request feedback and how to use it, CAFE trains the agent to recover from its own failures and shapes rewards both online and offline. Experiments on seven search benchmarks show that CAFE outperforms other RL‑based agents, maintains gains on out‑of‑domain tests, and reduces hallucinations, indicating that co‑evolving feedback is essential for self‑improving search agents.
The paper introduces SkillPivot, a framework that guides large language model agents to evolve their skills by pinpointing the exact moment a useful problem‑solving sequence turns into an erroneous suffix. SkillPivot uses execution validity, goal progress, and action diversity to detect this deviation point, then employs a stronger teacher to generate a successful alternative from the same prefix. By contrasting the failed and successful suffixes, the method produces localized, compact skill updates that preserve existing effective guidance, outperforming other skill‑evolution techniques on benchmarks such as ToolQA, LogicBench, and WildClawBench.
arXiv:2601. 07055v2 Announce Type: replace Abstract: As high-quality data becomes increasingly difficult to obtain, self-evolution without curated training data has emerged as a promising paradigm.
Outcome-supervised search agents learn when and how to retrieve evidence, but terminal rewards neither localize intermediate errors nor redirect an ongoing trajectory before those errors compound. Tre...
Although agent skills equip LLMs with reusable procedural knowledge, manual maintenance suffers from high costs, unscalability, and misalignment. Real-world deployments thus require autonomous, on-demand skill evolution at test time, constrained by limited interaction budgets and a lack of training or validation sets.
The paper introduces Traverse, a benchmark of 2,518 agent trajectories and 6,967 annotated mistakes across software engineering, computer use, and science tasks, revealing that failures often go unrecovered and can cause irreversible harm before a run is deemed successful. It shows that human judges struggle to detect the first mistake in most runs, while a 4‑billion‑parameter verifier called Scout can locate failures more effectively and improve task success when used to select among candidate runs. The study demonstrates that making failure detection inexpensive and reliable can enable long‑horizon agents to learn from their own mistakes and increase trustworthiness in autonomous AI.