Solving Math Word Problems with Process-based and Outcome-based Feedback

Jonathan Uesato, Nate Kushman, Ramana Kumar, H. Francis Song, Noah Yamamoto Siegel, Lisa Wang, Antonia Creswell, Geoffrey Irving, Irina Higgins

Keywords: language models, reasoning, reward models

Abstract: Recent work has shown that prompting language models to generate reasoning steps improves performance on many reasoning tasks. When moving beyond prompting, this raises the question of how we should supervise the finetuning of such models: outcome-based approaches which supervise the final result, or process-based approaches which supervise the reasoning process itself? Differences between these approaches might naturally be expected not just in final-answer errors but also in reasoning errors, which can be difficult to detect and are problematic in many real-world domains such as education. We run the first comprehensive comparison between process- and outcome-based approaches trained on a natural language task, GSM8K. We find that pure outcome-based supervision produces similar final-answer error rates with less label supervision. However, for correct reasoning steps we find it necessary to use process-based supervision or supervision from learned reward models that emulate process-based feedback. In total, we improve the previous best results from 16.8% $\rightarrow$ 12.7% final-answer error and 14.0% $\rightarrow$ 3.4% reasoning error among final-answer-correct solutions.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Submission Guidelines: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning

TL;DR: Both process- and outcome-based feedback with all the tricks achieve similar final-answer error rates and SOTA results, but generating accurate reasoning steps requires either process-based supervision, or a reward model that emulates it.