arXiv cs.AI / cs.LG / cs.CL·18d agoMeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM Agents#agents#benchmark#llm-agentsAI research