Federation over Text: Insight Sharing for Multi-Agent Reasoning

TL;DR

The paper introduces Federation over Text (FoT), a novel framework enabling multi-agent reasoning and insight sharing at the semantic level to improve performance across various tasks without gradient-based training.

Contribution

FoT is the first to facilitate semantic-level federated reasoning among agents, creating a shared insight library that enhances multi-task and cross-domain problem solving.

Findings

FoT improves downstream task accuracy by 24%.

FoT reduces reasoning tokens by 28%.

FoT generates insights covering over 90% of major contributions in research papers.

Abstract

LLM-powered agents often reason from scratch when presented with a new problem instance and lack automatic mechanisms to transfer learned skills to other agents. We propose a federated learning-like framework, Federation over Text (FoT), that enables multiple agents solving different tasks to collectively generate a shared library of metacognitive insights by iteratively federating their local reasoning processes. Instead of federation over gradients (e.g., as in distributed training), FoT operates at the semantic level without any gradient optimization or supervision signal. Iteratively, each agent does local thinking and self-improvement on their specific tasks independently, and shares reasoning traces with a central server, which aggregates and distills them into a cross-task (and cross-domain) insight library that existing and future agents can leverage to improve performance on related tasks. Experiments show that FoT improves reasoning effectiveness and efficiency across a wide range of challenging applications, including mathematical problem solving, cross-domain collaboration, and machine learning research insight discovery. Specifically, it improves average accuracies of downstream tasks by 24% while reducing the reasoning tokens by 28% across the first two applications. In the research insight discovery application, FoT is able to generate insights that cover over 90% of the major contributions in the subsequent papers.