Two-way Evidence self-Alignment based Dual-Gated Reasoning Enhancement

2025/05/22 by Kexin Zhang, Zhang, Kexin, Junlan Chen +11
Computer Science · #Computation and Language (cs.CL) #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Multimodal Machine Learning Applications #Topic Modeling

paper · pdf · doi:10.48550/arxiv.2505.16806

openalex publication_date 2025/05/22 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

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.

Citations

• Advancing Reasoning in Large Language Models: Promising Methods and Approaches
• Search-o1: Agentic Search-Enhanced Large Reasoning Models
• ChunkRAG: Novel LLM-Chunk Filtering Method for RAG Systems
• On the Reliability of Large Language Models for Causal Discovery
• Adaptive-RAG: Learning to Adapt Retrieval-Augmented Large Language Models through Question Complexity
• RA-ISF: Learning to Answer and Understand from Retrieval Augmentation via Iterative Self-Feedback
• BioXP-0.5B: Explainable Medical-AI via RL-GRPO
• Multistage Collaborative Knowledge Distillation from a Large Language Model for Semi-Supervised Sequence Generation
• mPLUG-Owl2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration
• Self-prompted Chain-of-Thought on Large Language Models for Open-domain Multi-hop Reasoning
• On the Over-Memorization During Natural, Robust and Catastrophic Overfitting
• Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes
• G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment
• Large Language Models Can Be Easily Distracted by Irrelevant Context
• Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions
• Measuring and Narrowing the Compositionality Gap in Language Models
• Automatic Chain of Thought Prompting in Large Language Models
• Least-to-Most Prompting Enables Complex Reasoning in Large Language Models
• Entailment Tree Explanations via Iterative Retrieval-Generation Reasoner
• Generated Knowledge Prompting for Commonsense Reasoning
• Learning Dense Representations of Phrases at Scale
• Constructing A Multi-hop QA Dataset for Comprehensive Evaluation of Reasoning Steps
• Learning to Explain: Datasets and Models for Identifying Valid Reasoning\n Chains in Multihop Question-Answering
• Leveraging Passage Retrieval with Generative Models for Open Domain\n Question Answering
• Language Models are Few-Shot Learners
• Dense Passage Retrieval for Open-Domain Question Answering
• Unsupervised Question Decomposition for Question Answering
• REALM: Retrieval-Augmented Language Model Pre-Training
• Break It Down: A Question Understanding Benchmark
• Multi-hop Reading Comprehension through Question Decomposition and Rescoring
• Interpretable Neural Predictions with Differentiable Binary Variables
• HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering
• Implicit Reparameterization Gradients

Related

• MR-Align: Meta-Reasoning Informed Factuality Alignment for Large Reasoning Models
• Self-Training Meets Consistency: Improving LLMs' Reasoning with Consistency-Driven Rationale Evaluation
• Self-Contradictory Reasoning Evaluation and Detection
• Making Reasoning Matter: Measuring and Improving Faithfulness of Chain-of-Thought Reasoning
• Anchored Alignment for Self-Explanations Enhancement