arXiv cs.AIWednesday · May 27, 2026FREE

LELA: An End-to-end LLM-based Entity Linking Framework with Zero-shot Domain Adaptation

entity-linkingnerllmpython-libraryzero-shot

LELA, introduced in a paper on arXiv (2605.26956v1), extends a modular LLM-based entity disambiguation method into a practical Python library. The library integrates zero-shot Named Entity Recognition (NER) to provide a complete end-to-end pipeline for entity linking. Unlike existing approaches that are tied to specific knowledge bases and domains, LELA is domain-agnostic and requires no fine-tuning or domain-specific training. The paper presents experimental results validating LELA's performance and robustness across diverse entity linking settings. A demo is available for users to test the system on their own input texts. The library is designed for real-world usage, making entity linking more accessible for developers working with various domains and knowledge bases.

// why it matters

Developers can now perform entity linking without domain-specific training, reducing setup time and improving portability.

Sources

Primary · arXiv cs.AI

▸ Read original at arxiv.org

Cross-Entropy Games and Frost Training Hierarchical Prompt-Domain Control and Learning for Resource-Constrained Agentic Language Models When Context Flips, Safety Breaks: Diagnosing Brittle Safety in Aligned Language Models

LELA: An End-to-end LLM-based Entity Linking Framework with Zero-shot Domain Adaptation

Sources

Related

Like this? Get the next digest.