English

Reverse Thinking Makes LLMs Stronger Reasoners

Computation and Language 2025-03-11 v2 Artificial Intelligence Machine Learning

Abstract

Reverse thinking plays a crucial role in human reasoning. Humans can reason not only from a problem to a solution but also in reverse, i.e., start from the solution and reason towards the problem. This often enhances overall reasoning performance as it enables consistency checks between their forward and backward thinking. To enable Large Language Models (LLMs) to perform reverse thinking, we introduce Reverse-Enhanced Thinking (RevThink), a framework composed of data augmentation and learning objectives. In RevThink, we augment the dataset by collecting structured forward-backward reasoning from a teacher model, consisting of: (1) the original question, (2) forward reasoning, (3) backward question, and (4) backward reasoning. We then employ three objectives to train a smaller student model in a multi-task learning fashion: (a) generate forward reasoning from a question, (b) generate a backward question from a question, and (c) generate backward reasoning from the backward question. Experiments across 12 datasets covering commonsense, math, and logical reasoning show an average 13.53% improvement over the student model's zero-shot performance and a 6.84% improvement over the strongest knowledge distillation baselines. Moreover, our method demonstrates sample efficiency -- using only 10% of the correct forward reasoning from the training data, it outperforms a standard fine-tuning method trained on 10x more forward reasoning. RevThink also exhibits strong generalization to out-of-distribution held-out datasets.

Keywords

Cite

@article{arxiv.2411.19865,
  title  = {Reverse Thinking Makes LLMs Stronger Reasoners},
  author = {Justin Chih-Yao Chen and Zifeng Wang and Hamid Palangi and Rujun Han and Sayna Ebrahimi and Long Le and Vincent Perot and Swaroop Mishra and Mohit Bansal and Chen-Yu Lee and Tomas Pfister},
  journal= {arXiv preprint arXiv:2411.19865},
  year   = {2025}
}

Comments

Accepted to NAACL 2025

R2 v1 2026-06-28T20:17:07.273Z