Large Language Model Cascades with Mixture of Thoughts Representations for Cost-efficient Reasoning

TL;DR

This paper introduces an LLM cascade approach that uses answer consistency and mixture of thought representations to reduce reasoning task costs while maintaining high performance, by selectively deploying stronger models only when needed.

Contribution

It proposes a novel cascade pipeline utilizing answer consistency and mixture of thought representations to efficiently allocate LLM resources for reasoning tasks.

Findings

Achieves comparable performance to strong LLMs at 40% of the cost.

Uses answer consistency as a signal for question difficulty.

Demonstrates effectiveness on six reasoning benchmark datasets.

Abstract

Large language models (LLMs) such as GPT-4 have exhibited remarkable performance in a variety of tasks, but this strong performance often comes with the high expense of using paid API services. In this paper, we are motivated to study building an LLM cascade to save the cost of using LLMs, particularly for performing reasoning (e.g., mathematical, causal) tasks. Our cascade pipeline follows the intuition that simpler questions can be addressed by a weaker but more affordable LLM, whereas only the challenging questions necessitate the stronger and more expensive LLM. To realize this decision-making, we consider the "answer consistency" of the weaker LLM as a signal of the question difficulty and propose several methods for the answer sampling and consistency checking, including one leveraging a mixture of two thought representations (i.e., Chain-of-Thought and Program-of-Thought). Through experiments on six reasoning benchmark datasets, with GPT-3.5-turbo and GPT-4 being the weaker and stronger LLMs, respectively, we demonstrate that our proposed LLM cascades can achieve performance comparable to using solely the stronger LLM but require only 40% of its cost.