SciAgent: Tool-augmented Language Models for Scientific Reasoning

TL;DR

This paper introduces SciAgent, a tool-augmented framework for scientific reasoning that enhances LLMs' problem-solving by integrating scalable tools, demonstrated through a new dataset and benchmark across multiple scientific domains.

Contribution

The paper presents a novel task setting for tool-augmented scientific reasoning, a large dataset MathFunc, and a new model SciAgent that effectively uses tools for scientific problem solving.

Findings

SciAgent-Mistral-7B outperforms same-size LLMs by over 13% accuracy.

SciAgent-DeepMath-7B significantly surpasses ChatGPT in scientific reasoning.

The approach improves LLMs' ability to utilize tools for complex scientific tasks.

Abstract

Scientific reasoning poses an excessive challenge for even the most advanced Large Language Models (LLMs). To make this task more practical and solvable for LLMs, we introduce a new task setting named tool-augmented scientific reasoning. This setting supplements LLMs with scalable toolsets, and shifts the focus from pursuing an omniscient problem solver to a proficient tool-user. To facilitate the research of such setting, we construct a tool-augmented training corpus named MathFunc which encompasses over 30,000 samples and roughly 6,000 tools. Building on MathFunc, we develop SciAgent to retrieve, understand and, if necessary, use tools for scientific problem solving. Additionally, we craft a benchmark, SciToolBench, spanning five scientific domains to evaluate LLMs' abilities with tool assistance. Extensive experiments on SciToolBench confirm the effectiveness of SciAgent. Notably, SciAgent-Mistral-7B surpasses other LLMs with the same size by more than 13% in absolute accuracy. Furthermore, SciAgent-DeepMath-7B shows much superior performance than ChatGPT.