Making Reasoning Matter: Measuring and Improving Faithfulness of Chain-of-Thought Reasoning

TL;DR

This paper investigates the faithfulness of reasoning steps in large language models and introduces FRODO, a framework that improves the accuracy and robustness of reasoning by training models to generate and rely on correct intermediate inferences.

Contribution

The paper presents FRODO, a novel framework that enhances reasoning faithfulness and robustness in language models through causal and counterfactual training methods.

Findings

FRODO outperforms four baselines in reasoning tasks.

FRODO improves out-of-distribution generalization.

FRODO's rationales are more faithful to final answers.

Abstract

Large language models (LLMs) have been shown to perform better when asked to reason step-by-step before answering a question. However, it is unclear to what degree the model's final answer is faithful to the stated reasoning steps. In this paper, we perform a causal mediation analysis on twelve LLMs to examine how intermediate reasoning steps generated by the LLM influence the final outcome and find that LLMs do not reliably use their intermediate reasoning steps when generating an answer. To address this issue, we introduce FRODO, a framework to tailor small-sized LMs to generate correct reasoning steps and robustly reason over these steps. FRODO consists of an inference module that learns to generate correct reasoning steps using an implicit causal reward function and a reasoning module that learns to faithfully reason over these intermediate inferences using a counterfactual and causal preference objective. Our experiments show that FRODO significantly outperforms four competitive baselines. Furthermore, FRODO improves the robustness and generalization ability of the reasoning LM, yielding higher performance on out-of-distribution test sets. Finally, we find that FRODO's rationales are more faithful to its final answer predictions than standard supervised fine-tuning.