Adaptive Multi-Expert Reasoning via Difficulty-Aware Routing and Uncertainty-Guided Aggregation

TL;DR

This paper introduces Adaptive Multi-Expert Reasoning (AMR), a framework that improves math reasoning in large language models by dynamically routing problems based on difficulty and aggregating responses with uncertainty guidance.

Contribution

The paper presents a novel adaptive reasoning framework with difficulty-aware routing and uncertainty-guided aggregation, enhancing LLM performance on math benchmarks without additional training data.

Findings

AMR achieved 75.28% accuracy on GSM8K with only original training data.

AMR outperformed most 7B models trained on synthetic data.

The framework effectively manages problem difficulty and response uncertainty.

Abstract

Large language models (LLMs) demonstrate strong performance in math reasoning benchmarks, but their performance varies inconsistently across problems with varying levels of difficulty. This paper describes Adaptive Multi-Expert Reasoning (AMR), a framework that focuses on problem complexity by reasoning with dynamically adapted strategies. An agile routing system that focuses on problem text predicts problems' difficulty and uncertainty and guides a reconfigurable sampling mechanism to manage the breadth of generation. Three specialized experts create candidate responses, which are modified during multiple correction and finalization phases. A neural verifier assesses the correctness of responses, while a clustering-based aggregation technique identifies the final candidate answer based on a combination of consensus and answer quality. When evaluated on the GSM8K dataset, AMR achieved 75.28% accuracy while only using the original training data. This result outperformed the majority of comparable 7B models that were trained on synthetic data. This showcases that models using difficulty-based routing and uncertainty-driven aggregation are efficient and effective in improving math reasoning models' robustness.