Year: 2026
Type: conference-paper
Abstract: Diffusion models have introduced a generative paradigm for Named Entity Recognition (NER), formulating the task as refining entity spans from noise.While promising, our analysis on the ACE2004 dataset reveals a limitation when training with Exponential Moving Average (EMA): the model performance often peaks at a single inference step (γ = 1) and plateaus or degrades with additional steps.This suggests that under standard stable training configurations, the model may function primarily as a one-step generator rather than leveraging the iterative refinement capability characteristic of diffusion models.To address this, we propose an Initializer-Restorer approach.Instead of initializing the reverse process from random Gaussian noise, we utilize a preliminary set of candidate spans generated by a standard NER model (e.g., BERT or GLiNER).This allows the diffusion model to start from an informed, diverse prior, enabling effective iterative restoration.We investigate different training strategies for the restorer and find that a hybrid strategy mixing ground truth and noisy predictions is essential.Experiments on ACE2004, GENIA, and CleanCoNLL show that our approach improves performance over the baseline, particularly when multiple restoration steps are employed.For instance, on CleanCoNLL, our method achieves an F1 score of 94.70%, compared to 93.79% for the baseline.
Authors: Long Hai Trieu, Phí Minh Hieu, Makoto Miwa
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7166854384