A dynamic parsing and relationship inference model for entity features that incorporates gating units and pooling strategies to dynamically update hidden states is proposed that outperforms the baseline model Adaptive Thresholding and Localized Context Pooling, particularly in extracting potential relationships across sentence–boundary entity pairs.
A three-stage framework made up of neighborhood-aware retrieval, dual reranking, and score fusion that achieves the state of the art on average across five widely-used benchmarks and remains efficient at inference.
Document-level relation extraction (DocRE) aims to extract relations among multiple entities across extended contexts while maintaining consistency across predicted triples. Although large language models (LLMs) show remarkable reasoning capabilities in information extraction, their predictions are typically generated independently for each candidate triple and may violate fundamental relational constraints such as transitivity, symmetry, and functional uniqueness, leading to contradictory and unreliable outputs. We propose CONSISTRE, a unified consistency-aware framework for DocRE that addresses this limitation through two complementary tracks. The first operates at inference time for black-box LLMs, combining constraint-aware prompting, constraint-based verification, and iterative self-reflection to refine predictions without task-specific fine-tuning. The second injects consistency knowledge into smaller open-source models via a knowledge distillation and reinforcement learning pipeline: reasoning traces from a powerful teacher are distilled into a student via supervised fine-tuning, followed by GRPO alignment using a composite reward that jointly optimizes extraction performance and relational consistency. Together, the two tracks cover both API-accessible and locally deployable scenarios under a unified consistency formulation. Experiments on DocRED show that both tracks outperform their baselines, with the inference-time track achieving competitive F1 using off-the-shelf black-box LLMs and the training-time track substantially narrowing the gap between 7--8B open-source models and state-of-the-art proprietary LLMs at a fraction of their inference cost. Ablation studies confirm that explicit consistency modeling mitigates relational contradictions and enhances the reliability of LLM-based DocRE across both deployment paradigms.
This work introduces an ontology-driven framework to quantify and enforce structural consistency in DocRE datasets and reveals substantial structural noise in DocRED distant and demonstrates that such inconsistencies propagate to model predictions.
Laura Menotti, Stefano Marchesin, Gianmaria Silvello· 0 citations
Named Entity Recognition (NER) extracts structured semantic elements, such as persons, organizations, or biomedical concepts, that enhance document understanding. Rather than replacing pretrained document embeddings, we refine them by injecting entity-level relational information. By constructing a document graph based on named-entity similarity, we apply spectral filtering to smooth the embedding matrix, yielding entity-aware representations compatible with any clustering algorithm. This framework is model-agnostic and lightweight. Furthermore, it can distill a set of high-confidence Must-Link (ML) constraints, pairs of documents that should be assigned to the same cluster, to guide constrained clustering. Experimental results on benchmark datasets demonstrate that our approach significantly improves clustering performance, highlighting the value of entity-aware smoothing over standard pretrained representations.
Imed Keraghel, Mohamed Nadif· Annual International ACM SIG...· 0 citations
This analysis shows that, in new sentences, the contextual associations of tokens representing old entity types exhibit a significantly stronger bias towards new entity types compared to their contexts in old sentences, which intensifies the degradation of old knowledge while promoting the overfitting of new knowledge.
Duzhen Zhang, Yahan Yu, Xiuyi Chen et al.· IEEE Transactions on Artific...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
