ImCoref-CeS: An Improved Lightweight Pipeline for Coreference Resolution with LLM-based Checker-Splitter Refinement

2025-10-11

AI Summary

Coreference resolution faces a critical challenge in effectively integrating supervised small-model approaches with the reasoning capabilities of large language models (LLMs). This paper proposes a lightweight fusion framework: first, a supervised detection-clustering pipeline built upon small language models, enhanced with a bridging module for long-text encoding and a dual-affine scoring mechanism to model positional relations between mentions; second, an LLM deployed as a multi-role “Checker-Splitter” agent—without fine-tuning—that automatically validates mention validity and detects/splits erroneous clusters via prompt engineering. The framework enables efficient, LLM-augmented post-processing while preserving the strengths of supervised modeling. It achieves significant improvements over state-of-the-art methods on multiple benchmarks, including OntoNotes and LitBank, demonstrating both the feasibility and superiority of synergistic optimization between supervised models and LLM-based reasoning.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Coreference Resolution (CR) is a critical task in Natural Language Processing (NLP). Current research faces a key dilemma: whether to further explore the potential of supervised neural methods based on small language models, whose detect-then-cluster pipeline still delivers top performance, or embrace the powerful capabilities of Large Language Models (LLMs). However, effectively combining their strengths remains underexplored. To this end, we propose extbf{ImCoref-CeS}, a novel framework that integrates an enhanced supervised model with LLM-based reasoning. First, we present an improved CR method ( extbf{ImCoref}) to push the performance boundaries of the supervised neural method by introducing a lightweight bridging module to enhance long-text encoding capability, devising a biaffine scorer to comprehensively capture positional information, and invoking a hybrid mention regularization to improve training efficiency. Importantly, we employ an LLM acting as a multi-role Checker-Splitter agent to validate candidate mentions (filtering out invalid ones) and coreference results (splitting erroneous clusters) predicted by ImCoref. Extensive experiments demonstrate the effectiveness of ImCoref-CeS, which achieves superior performance compared to existing state-of-the-art (SOTA) methods.

Problem

• Improving coreference resolution by combining supervised models with LLM reasoning
• Enhancing long-text encoding and positional information capture in CR
• Validating and refining coreference clusters using LLM-based checker-splitter

Innovation

• Enhanced supervised model with lightweight bridging module
• Biaffine scorer capturing comprehensive positional information
• LLM-based Checker-Splitter agent refining coreference results