Multi-Agent Sampling: Scaling Inference Compute for Data Synthesis with Tree Search-Based Agentic Collaboration

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Ye, Hai, Lin, Mingbao, Ng, Hwee Tou, Yan, Shuicheng
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918024841265152
author Ye, Hai
Lin, Mingbao
Ng, Hwee Tou
Yan, Shuicheng
author_facet Ye, Hai
Lin, Mingbao
Ng, Hwee Tou
Yan, Shuicheng
contents Scaling laws for inference compute in multi-agent systems remain under-explored compared to single-agent scenarios. This work aims to bridge this gap by investigating the problem of data synthesis through multi-agent sampling, where synthetic responses are generated by sampling from multiple distinct language models. Effective model coordination is crucial for successful multi-agent collaboration. Unlike previous approaches that rely on fixed workflows, we treat model coordination as a multi-step decision-making process, optimizing generation structures dynamically for each input question. We introduce Tree Search-based Orchestrated Agents~(TOA), where the workflow evolves iteratively during the sequential sampling process. To achieve this, we leverage Monte Carlo Tree Search (MCTS), integrating a reward model to provide real-time feedback and accelerate exploration. Our experiments on alignment, machine translation, and mathematical reasoning demonstrate that multi-agent sampling significantly outperforms single-agent sampling as inference compute scales. TOA is the most compute-efficient approach, achieving SOTA performance on WMT and a 72.2\% LC win rate on AlpacaEval. Moreover, fine-tuning with our synthesized alignment data surpasses strong preference learning methods on challenging benchmarks such as Arena-Hard and AlpacaEval.
format Preprint
id arxiv_https___arxiv_org_abs_2412_17061
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Multi-Agent Sampling: Scaling Inference Compute for Data Synthesis with Tree Search-Based Agentic Collaboration
Ye, Hai
Lin, Mingbao
Ng, Hwee Tou
Yan, Shuicheng
Computation and Language
Scaling laws for inference compute in multi-agent systems remain under-explored compared to single-agent scenarios. This work aims to bridge this gap by investigating the problem of data synthesis through multi-agent sampling, where synthetic responses are generated by sampling from multiple distinct language models. Effective model coordination is crucial for successful multi-agent collaboration. Unlike previous approaches that rely on fixed workflows, we treat model coordination as a multi-step decision-making process, optimizing generation structures dynamically for each input question. We introduce Tree Search-based Orchestrated Agents~(TOA), where the workflow evolves iteratively during the sequential sampling process. To achieve this, we leverage Monte Carlo Tree Search (MCTS), integrating a reward model to provide real-time feedback and accelerate exploration. Our experiments on alignment, machine translation, and mathematical reasoning demonstrate that multi-agent sampling significantly outperforms single-agent sampling as inference compute scales. TOA is the most compute-efficient approach, achieving SOTA performance on WMT and a 72.2\% LC win rate on AlpacaEval. Moreover, fine-tuning with our synthesized alignment data surpasses strong preference learning methods on challenging benchmarks such as Arena-Hard and AlpacaEval.
title Multi-Agent Sampling: Scaling Inference Compute for Data Synthesis with Tree Search-Based Agentic Collaboration
topic Computation and Language
url https://arxiv.org/abs/2412.17061