Logic-Oriented Retriever Enhancement via Contrastive Learning

2026-02-01

AI Summary

This work addresses the limitation of existing retrievers that rely on surface-level similarity when handling complex logical queries, which constrains the performance of large language models in knowledge-intensive tasks. To overcome this, the authors propose LORE, a method that leverages fine-grained contrastive learning to activate the model’s inherent logical reasoning capabilities, steering embedding representations to align with deep logical structures rather than shallow semantic cues. LORE requires no external supervision, additional resources, or pre-retrieval analysis, while preserving index compatibility and computational efficiency. Experimental results demonstrate that LORE significantly improves both retrieval accuracy and downstream generation quality across multiple knowledge-intensive benchmarks. The code and datasets are publicly released.

Technology Category

• Knowledge Representation and Reasoning: Logic Programming
• Machine Learning: Statistical Relational/Logic Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

Large language models (LLMs) struggle in knowledge-intensive tasks, as retrievers often overfit to surface similarity and fail on queries involving complex logical relations. The capacity for logical analysis is inherent in model representations but remains underutilized in standard training. LORE (Logic ORiented Retriever Enhancement) introduces fine-grained contrastive learning to activate this latent capacity, guiding embeddings toward evidence aligned with logical structure rather than shallow similarity. LORE requires no external upervision, resources, or pre-retrieval analysis, remains index-compatible, and consistently improves retrieval utility and downstream generation while maintaining efficiency. The datasets and code are publicly available at https://github.com/mazehart/Lore-RAG.

Problem

• retrieval
• logical reasoning
• knowledge-intensive tasks
• large language models
• contrastive learning

Innovation

• contrastive learning
• logic-oriented retrieval
• retriever enhancement
• large language models
• knowledge-intensive tasks