A Toolbox, Not a Hammer -- Multi-TAG: Scaling Math Reasoning with Multi-Tool Aggregation

TL;DR

Multi-TAG introduces a multi-tool aggregation framework that enables large language models to invoke and combine multiple tools simultaneously during math reasoning, significantly improving accuracy on complex benchmarks without fine-tuning.

Contribution

It proposes a novel, fine-tuning-free framework allowing LLMs to use multiple tools concurrently for complex math reasoning, outperforming existing single-tool methods.

Findings

Outperforms state-of-the-art baselines by 6.0% to 7.5% on multiple benchmarks.

Works with both open-weight and proprietary LLMs without fine-tuning.

Enhances reasoning robustness and accuracy through multi-tool aggregation.

Abstract

Augmenting large language models (LLMs) with external tools is a promising avenue for developing high-performance mathematical reasoning systems. Prior tool-augmented approaches typically finetune an LLM to select and invoke a single tool at each reasoning step and show promising results on simpler math reasoning benchmarks such as GSM8K. However, these approaches struggle with more complex math problems that require precise reasoning over multiple steps. To address this limitation, in this work, we propose Multi-TAG, a Multi-Tool AGgregation-based framework. Instead of relying on a single tool, Multi-TAG guides an LLM to concurrently invoke multiple tools at each reasoning step. It then aggregates their diverse outputs to verify and refine the reasoning process, enhancing solution robustness and accuracy. Notably, Multi-TAG is a finetuning-free, inference-only framework, making it readily applicable to any LLM backbone, including large open-weight models which are computationally expensive to finetune and proprietary frontier models which cannot be finetuned with custom recipes. We evaluate Multi-TAG on four challenging benchmarks: MATH500, AIME, AMC, and OlympiadBench. Across both open-weight and closed-source LLM backbones, Multi-TAG consistently and substantially outperforms state-of-the-art baselines, achieving average improvements of 6.0% to 7.5% over state-of-the-art baselines.