Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Learning to Decode Collaboratively with Multiple Language Models

About

We propose a method to teach multiple large language models (LLM) to collaborate by interleaving their generations at the token level. We model the decision of which LLM generates the next token as a latent variable. By optimizing the marginal likelihood of a training set under our latent variable model, the base LLM automatically learns when to generate itself and when to call on one of the ``assistant'' language models to generate, all without direct supervision. Token-level collaboration during decoding allows for a fusion of each model's expertise in a manner tailored to the specific task at hand. Our collaborative decoding is especially useful in cross-domain settings where a generalist base LLM learns to invoke domain expert models. On instruction-following, domain-specific QA, and reasoning tasks, we show that the performance of the joint system exceeds that of the individual models. Through qualitative analysis of the learned latent decisions, we show models trained with our method exhibit several interesting collaboration patterns, e.g., template-filling. Our code is available at https://github.com/clinicalml/co-llm.

Shannon Zejiang Shen, Hunter Lang, Bailin Wang, Yoon Kim, David Sontag• 2024

Related benchmarks

TaskDatasetResultRank
Mathematical ReasoningGSM8K
Accuracy45
212
Instruction FollowingAlpacaEval
Win Rate71.9
125
Biomedical Question AnsweringBioASQ
Factoid Acc24.7
11
Math ReasoningMATH (test)
Accuracy25
7
Showing 4 of 4 rows

Other info

Code

Follow for update