Stream of Search (SoS): Learning to Search in Language

TL;DR

This paper introduces a method called Stream of Search (SoS) that trains language models to learn search strategies in problem-solving tasks, significantly improving their accuracy and ability to solve previously unsolvable problems.

Contribution

It presents a unified language for representing search processes, pretraining techniques that enhance search accuracy, and policy improvement methods that enable models to self-improve and discover new strategies.

Findings

SoS pretraining increases search accuracy by 25%.

Models solve 36% more problems, including previously unsolvable ones.

Language models can learn, adapt, and discover search strategies.

Abstract

Language models are rarely shown fruitful mistakes while training. They then struggle to look beyond the next token, suffering from a snowballing of errors and struggling to predict the consequence of their actions several steps ahead. In this paper, we show how language models can be taught to search by representing the process of search in language, as a flattened string -- a stream of search (SoS). We propose a unified language for search that captures an array of different symbolic search strategies. We demonstrate our approach using the simple yet difficult game of Countdown, where the goal is to combine input numbers with arithmetic operations to reach a target number. We pretrain a transformer-based language model from scratch on a dataset of streams of search generated by heuristic solvers. We find that SoS pretraining increases search accuracy by 25% over models trained to predict only the optimal search trajectory. We further finetune this model with two policy improvement methods: Advantage-Induced Policy Alignment (APA) and Self-Taught Reasoner (STaR). The finetuned SoS models solve 36% of previously unsolved problems, including problems that cannot be solved by any of the heuristic solvers. Our results indicate that language models can learn to solve problems via search, self-improve to flexibly use different search strategies, and potentially discover new ones.