aiqbench analysis
This research introduces DE-NER, a framework that leverages dialogue elicitation to overcome prompt engineering limitations in zero-shot Named Entity Recognition tasks using Large Language Models. By fully extracting latent knowledge through conversational interactions with LLMs, the method achieves an average 3.75% F1 improvement over competitive baselines across multiple benchmarks without requiring extensive human intervention.
DE-NER : Zero-shot Named Entity Recognition via Dialogue Elicitation of Large Language Models
arXiv:2608.00538v1 Announce Type: new Abstract: Recent advancements of zero-shot Named Entity Recognition (NER) establish strong baselines by formulating sequence labeling into question answering where Large Language Models (LLMs) can be naturally adopted. However, existing LLM-based zero-shot NER methods suffer from the limitations of prompt and demonstration engineering. To address these issues with minimal human interventions, we introduce DE-NER, a dialogue elicitation framework which elicits the chatting ability of LLMs to fully extract the knowledge encoded in LLMs. Our experiments demonstrate that the proposed method outperform the competitive baselines in zero-shot settings across multiple benchmarks, with an average improvement of 3.75\% F1 points. Codes are released in https://github.com/kkkenshi/DE-NER.
Abstract:Recent advancements of zero-shot Named Entity Recognition (NER) establish strong baselines by formulating sequence labeling into question answering where Large Language Models (LLMs) can be naturally adopted. However, existing LLM-based zero-shot NER methods suffer from the limitations of prompt and demonstration engineering. To address these issues with minimal human interventions, we introduce DE-NER, a dialogue elicitation framework which elicits the chatting ability of LLMs to fully extract the knowledge encoded in LLMs. Our experiments demonstrate that the proposed method outperform the competitive baselines in zero-shot settings across multiple benchmarks, with an average improvement of 3.75\% F1 points. Codes are released in this https URL.
Computation and Language (cs.CL)
arXiv:2608.00538 [cs.CL]
(or arXiv:2608.00538v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2608.00538 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 199 words
Related coverage
Just Pass Twice: Efficient Token Classification with LLMs for Zero-Shot NER
arXiv · Aug 27, 2026
Practical NLP in the Browser with Transformers.js
KDnuggets · May 29, 2026
Language Specificity vs. Domain Diversity: Benchmarking Transformers for Bangla Medical NER
arXiv · Sep 26, 2026
NE-R1: Enhancing Named Entity Recognition Model via Reinforcement Learning
arXiv · Sep 3, 2026