New method helps AI agents forget unhelpful memories to improve tasks

Original: MeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM Agents

Source: arxiv.org ↗