The paper introduces TASTE, an automatic task synthesis method that generates challenging agent benchmarks by evolving tool sequences, demonstrating that existing benchmarks are saturated and that TAS…

SkillRevise is an execution-grounded framework that iteratively refines initial, imperfect LLM agent skills by diagnosing defects from execution evidence and applying empirically validated edits, sign…

The paper introduces BenchJack, an automated red-teaming system that systematically audits popular AI agent benchmarks, revealing numerous reward-hacking exploits and demonstrating a method to signifi…

The paper proposes Meta-Team, an experience-driven framework that enables multi-agent systems (MAS) to collaboratively self-evolve by transforming complex execution experiences into reusable improveme…

The paper proposes HiSME, a lightweight hierarchical skill meta-evolving solution that jointly optimizes skills and the skill evolving strategy by learning meta-skills from task execution traces, lead…

The paper demonstrates that self-reflective agents can systematically confabulate incorrect memories, leading them to fail tasks even when the environment resets, and proposes a metric and mitigation…

The paper distinguishes between a model's ability to generate useful updates for external agent components (harness-updating) and its ability to benefit from those updates (harness-benefit), finding t…

The paper introduces Unified Context Evolution (UCE), a gradient-free framework that externalizes and manages agent experience into a typed, evolving library, significantly improving performance on mu…

The paper introduces Obsessive Experience Poisoning (OEP), a low-privilege black-box attack that poisons self-evolving LLM agents by generating locally correct but harmful experiences, causing dangero…

SCOPE introduces a data-free self-play framework that co-evolves a task-generating Challenger and a document-answering Solver, significantly improving open-ended performance on language models without…

SkillAdaptor is a novel, training-free framework that enables stable, step-level adaptation of external skills for LLM agents by precisely attributing failures to specific skills.

The paper introduces AGENTCL, a rigorous evaluation framework that uses controlled task streams to accurately measure an agent's ability to accumulate and reuse knowledge across multiple tasks, thereb…

The paper introduces CyberEvolver, a self-evolving agent framework that iteratively revises its own operational scaffold based on failed execution attempts, significantly improving cybersecurity agent…

The paper introduces LoopTrap, an automated red-teaming framework that demonstrates how malicious prompts can poison the termination judgment of LLM agents, causing unbounded computation.

TraceGraph introduces a graph-based framework to map agent decision-making across pooled trajectories, revealing hidden differences in agent behavior and improving performance by targeting known failu…

The paper investigates the limits of self-evolution in LLM reasoning under closed-loop settings, finding that while self-improvement is significant, it consistently falls short of perfect oracle super…

SkillSmith is a synergy-aware framework that jointly co-evolves skills and tools, significantly improving self-improving agent systems by modeling skill-tool interactions and diagnosing failures.

ExpGraph is a model-agnostic framework that uses a self-evolving experience graph to enable LLM agents to reuse past successful strategies and failure lessons, significantly improving performance acro…

The paper introduces Proteus, a self-evolving red-team framework that measures the adaptive leakage risk of LLM agent skills, demonstrating that current vetting methods significantly underestimate res…

The paper introduces MASA, a model-aware skill alignment framework that adaptively rewrites general and task-specific skills for LLM agents, achieving superior performance across diverse backbones and…