Search

直接匹配不多,已補上最新動態。

Tag: #arc1 results

ARC Learns Dynamic Agent Configurations

ARC Learns Dynamic Agent Configurations

ARC introduces a reinforcement learning policy to dynamically configure LLM-based agent systems per query, selecting optimal workflows, tools, and prompts. It outperforms fixed templates on reasoning and tool-augmented QA benchmarks. The approach boosts accuracy by up to 25% while cutting token and runtime costs.

ArXiv AIResearchFeb 13#research#arc#llm-agents