AI RESEARCH
Why Retrying Fails: Context Contamination in LLM Agent Pipelines
arXiv CS.AI
•
ArXi:2605.08563v1 Announce Type: new When an LLM agent fails a multi-step tool-augmented task and retries, the failed attempt typically remains in its context window -- contaminating the next attempt and elevating the per-step error rate beyond the base level. This context-contaminated restart phenomenon is widely observed in practice yet entirely lacks formal treatment. We