Model Swarms: Collaborative Search to Adapt LLM Experts via Swarm Intelligence

TL;DR

Model Swarms introduces a swarm intelligence-based collaborative search method for adapting large language models, enabling tuning-free, data-efficient, and flexible model adaptation across various tasks and domains.

Contribution

It presents a novel swarm intelligence algorithm for LLM adaptation that does not require tuning or specific expert assumptions, outperforming existing composition methods.

Findings

Achieves up to 21.0% improvement over baselines

Works effectively with as few as 200 examples

Enables discovery of new capabilities in LLM checkpoints

Abstract

We propose Model Swarms, a collaborative search algorithm to adapt LLMs via swarm intelligence, the collective behavior guiding individual systems. Specifically, Model Swarms starts with a pool of LLM experts and a utility function. Guided by the best-found checkpoints across models, diverse LLM experts collaboratively move in the weight space and optimize a utility function representing model adaptation objectives. Compared to existing model composition approaches, Model Swarms offers tuning-free model adaptation, works in low-data regimes with as few as 200 examples, and does not require assumptions about specific experts in the swarm or how they should be composed. Extensive experiments demonstrate that Model Swarms could flexibly adapt LLM experts to a single task, multi-task domains, reward models, as well as diverse human interests, improving over 12 model composition baselines by up to 21.0% across tasks and contexts. Further analysis reveals that LLM experts discover previously unseen capabilities in initial checkpoints and that Model Swarms enable the weak-to-strong transition of experts through the collaborative search process.