When Non-Adversarial Noise Hurts Learning: Benchmarking Skill Self-Evolution in Agents

arXiv:2610.19910 · cs.LG · Submitted 2026-10-14 · Read on arXiv

cs.LG

Submitted: 2026-10-14

Updated: 2026-10-14

Comments: 35 pages, 2 figures

Code: https://github.com/yutao1024/RSEClaw

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Adversarial Noise — 1× in this paper
• Benchmarking — 1× in this paper · explained in 4 episodes

Sources

• StressWeb: A Diagnostic Benchmark for Web Agent Robustness under Realistic Interaction Variability
• AgentPoison: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases
• From Untrusted Input to Trusted Memory: A Systematic Study of Memory Poisoning Attacks in LLM Agents
• Memory Injection Attacks on LLM Agents via Query-Only Interaction
• ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities?
• AttriMem: Attribution-Guided Process Feedback for Agent Memory Construction
• Rethinking Self-Evolving Agent Skills: Feedback Dynamics over Multiple Rounds
• Caution for the Environment: Multimodal LLM Agents are Susceptible to Environmental Distractions
• SpreadsheetBench: Towards Challenging Real World Spreadsheet Manipulation
• Practice Makes Unsafe: Skill Misevolution in Self-Improving LLM Agents
• DocVQA: A Dataset for VQA on Document Images
• Large Language Models Can Be Easily Distracted by Irrelevant Context
• Reflexion: Language Agents with Verbal Reinforcement Learning
• ALFWorld: Aligning Text and Embodied Environments for Interactive Learning
• Voyager: An Open-Ended Embodied Agent with Large Language Models
• AgentNoiseBench: Benchmarking Robustness of Tool-Using LLM Agents Under Noisy Condition
• Agent Workflow Memory
• Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory
• StreamBench: Towards Benchmarking Continuous Improvement of Language Agents
• MemEvoBench: Benchmarking Safety Risks from Memory Misevolution in LLM Agents

Related papers

• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
• AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
• Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
• Convergence issues in Relational Concept Analysis based on AOC-posets
• Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
• Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks