Saved in:
Bibliographic Details
Main Authors: Zhang, Kexin, Chen, Junlan, Li, Daifeng, Zhang, Yuxuan, Feng, Yangyang, Deng, Bowen, Chen, Weixu
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.16806
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916752509632512
author Zhang, Kexin
Chen, Junlan
Li, Daifeng
Zhang, Yuxuan
Feng, Yangyang
Deng, Bowen
Chen, Weixu
author_facet Zhang, Kexin
Chen, Junlan
Li, Daifeng
Zhang, Yuxuan
Feng, Yangyang
Deng, Bowen
Chen, Weixu
contents Large language models (LLMs) encounter difficulties in knowledge-intensive multi-step reasoning (KIMSR) tasks. One challenge is how to effectively extract and represent rationale evidence. The current methods often extract semantically relevant but logically irrelevant evidence, resulting in flawed reasoning and inaccurate responses. We propose a two-way evidence self-alignment (TW-ESA) module, which utilizes the mutual alignment between strict reasoning and LLM reasoning to enhance its understanding of the causal logic of evidence, thereby addressing the first challenge. Another challenge is how to utilize the rationale evidence and LLM's intrinsic knowledge for accurate reasoning when the evidence contains uncertainty. We propose a dual-gated reasoning enhancement (DGR) module to gradually fuse useful knowledge of LLM within strict reasoning, which can enable the model to perform accurate reasoning by focusing on causal elements in the evidence and exhibit greater robustness. The two modules are collaboratively trained in a unified framework ESA-DGR. Extensive experiments on three diverse and challenging KIMSR datasets reveal that ESA-DGR significantly surpasses state-of-the-art LLM-based fine-tuning methods, with remarkable average improvements of 4% in exact match (EM) and 5% in F1 score. The implementation code is available at https://anonymous.4open.science/r/ESA-DGR-2BF8.
format Preprint
id arxiv_https___arxiv_org_abs_2505_16806
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Two-way Evidence self-Alignment based Dual-Gated Reasoning Enhancement
Zhang, Kexin
Chen, Junlan
Li, Daifeng
Zhang, Yuxuan
Feng, Yangyang
Deng, Bowen
Chen, Weixu
Computation and Language
Information Retrieval
Large language models (LLMs) encounter difficulties in knowledge-intensive multi-step reasoning (KIMSR) tasks. One challenge is how to effectively extract and represent rationale evidence. The current methods often extract semantically relevant but logically irrelevant evidence, resulting in flawed reasoning and inaccurate responses. We propose a two-way evidence self-alignment (TW-ESA) module, which utilizes the mutual alignment between strict reasoning and LLM reasoning to enhance its understanding of the causal logic of evidence, thereby addressing the first challenge. Another challenge is how to utilize the rationale evidence and LLM's intrinsic knowledge for accurate reasoning when the evidence contains uncertainty. We propose a dual-gated reasoning enhancement (DGR) module to gradually fuse useful knowledge of LLM within strict reasoning, which can enable the model to perform accurate reasoning by focusing on causal elements in the evidence and exhibit greater robustness. The two modules are collaboratively trained in a unified framework ESA-DGR. Extensive experiments on three diverse and challenging KIMSR datasets reveal that ESA-DGR significantly surpasses state-of-the-art LLM-based fine-tuning methods, with remarkable average improvements of 4% in exact match (EM) and 5% in F1 score. The implementation code is available at https://anonymous.4open.science/r/ESA-DGR-2BF8.
title Two-way Evidence self-Alignment based Dual-Gated Reasoning Enhancement
topic Computation and Language
Information Retrieval
url https://arxiv.org/abs/2505.16806