MoA Mixture of Agents (2024年)（モーアミックスチャーオブエージェンツ）

概要

MoA Mixture of Agents は2024年6月Together AI + Duke University発表・Industry-leading multi-LLM ensemble 2024年Together AI MoA Mixture of Agents multi-LLM aggregation framework position確立。Industry-leading layered aggregator + beats GPT-4o on AlpacaEval 2.0 + open-source models only + 6 layers proposer + aggregator + Apache 2.0。

主な特徴・仕組み

• Industry-leading Together AI + Duke University (Junlin Wang + Jue Wang + Zhuoran Yang + Ce Zhang) 2024年6月 paper
• Industry-leading MoA Mixture of Agents multi-LLM aggregation framework signature
• Industry-leading multi-LLM ensemble signature (Qwen + LLaMA + Mixtral + WizardLM)
• Industry-leading layered aggregator signature (proposer + aggregator layers)
• Industry-leading beats GPT-4o on AlpacaEval 2.0 signature (65.1 vs 57.5)
• Industry-leading open-source models only signature (no proprietary needed)
• Industry-leading 6 layers proposer + aggregator architecture
• Industry-leading 2024年6月 ArXiv paper "Mixture-of-Agents Enhances Large Language Model Capabilities"
• Industry-leading Apache 2.0 license signature
• Industry-leading Together AI 2022 + Duke University collaboration

まとめ

MoA Mixture of Agents = 2024年6月Together AI + Duke University発表のpremium Apache 2.0 multi-LLM aggregation framework。Industry-leading multi-LLM ensemble Qwen + LLaMA + Mixtral + WizardLM + layered aggregator proposer + aggregator layers + beats GPT-4o on AlpacaEval 2.0 65.1 vs 57.5 + open-source models only no proprietary needed + 6 layers proposer + aggregator architecture + 2024年6月 ArXiv paper + Apache 2.0 + Together AI + Duke University Junlin Wang + Jue Wang + Zhuoran Yang + Ce Zhang 2024 collaboration position確立。