Self-Training Meets Consistency: Improving LLMs' Reasoning with Consistency-Driven Rationale Evaluation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lee, Jaehyeok, Sakaguchi, Keisuke, Bak, JinYeong
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915141160796160
author Lee, Jaehyeok
Sakaguchi, Keisuke
Bak, JinYeong
author_facet Lee, Jaehyeok
Sakaguchi, Keisuke
Bak, JinYeong
contents Self-training approach for large language models (LLMs) improves reasoning abilities by training the models on their self-generated rationales. Previous approaches have labeled rationales that produce correct answers for a given question as appropriate for training. However, a single measure risks misjudging rationale quality, leading the models to learn flawed reasoning patterns. To address this issue, we propose CREST (Consistency-driven Rationale Evaluation for Self-Training), a self-training framework that further evaluates each rationale through follow-up questions and leverages this evaluation to guide its training. Specifically, we introduce two methods: (1) filtering out rationales that frequently result in incorrect answers on follow-up questions and (2) preference learning based on mixed preferences from rationale evaluation results of both original and follow-up questions. Experiments on three question-answering datasets using open LLMs show that CREST not only improves the logical robustness and correctness of rationales but also improves reasoning abilities compared to previous self-training approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2411_06387
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Self-Training Meets Consistency: Improving LLMs' Reasoning with Consistency-Driven Rationale Evaluation
Lee, Jaehyeok
Sakaguchi, Keisuke
Bak, JinYeong
Machine Learning
Artificial Intelligence
Computation and Language
Self-training approach for large language models (LLMs) improves reasoning abilities by training the models on their self-generated rationales. Previous approaches have labeled rationales that produce correct answers for a given question as appropriate for training. However, a single measure risks misjudging rationale quality, leading the models to learn flawed reasoning patterns. To address this issue, we propose CREST (Consistency-driven Rationale Evaluation for Self-Training), a self-training framework that further evaluates each rationale through follow-up questions and leverages this evaluation to guide its training. Specifically, we introduce two methods: (1) filtering out rationales that frequently result in incorrect answers on follow-up questions and (2) preference learning based on mixed preferences from rationale evaluation results of both original and follow-up questions. Experiments on three question-answering datasets using open LLMs show that CREST not only improves the logical robustness and correctness of rationales but also improves reasoning abilities compared to previous self-training approaches.
title Self-Training Meets Consistency: Improving LLMs' Reasoning with Consistency-Driven Rationale Evaluation
topic Machine Learning
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2411.06387