Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.11133 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908828009758720 |
|---|---|
| author | Kohut, Zahar Shykula, Severyn Khamula, Dmytro Vysotskyi, Mykola Rumezhak, Taras Karpiv, Volodymyr |
| author_facet | Kohut, Zahar Shykula, Severyn Khamula, Dmytro Vysotskyi, Mykola Rumezhak, Taras Karpiv, Volodymyr |
| contents | Diffusion language models generate text through iterative refinement, a process that is often computationally inefficient because many tokens reach stability long before the final denoising step. We introduce a training-free, token-level early stopping approach that identifies convergence independently at each position. Our method leverages lightweight signals derived from the model's predictions and local context to dynamically determine when individual tokens can be finalized. This yields adaptive per-token freezing without task-specific fine-tuning, substantially reducing the total number of diffusion steps required. Across diverse benchmarks, spanning mathematical reasoning, general question answering, and scientific understanding, our approach achieves state-of-the-art efficiency gains while preserving generation quality. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_11133 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Just on Time: Token-Level Early Stopping for Diffusion Language Models Kohut, Zahar Shykula, Severyn Khamula, Dmytro Vysotskyi, Mykola Rumezhak, Taras Karpiv, Volodymyr Machine Learning Computation and Language Diffusion language models generate text through iterative refinement, a process that is often computationally inefficient because many tokens reach stability long before the final denoising step. We introduce a training-free, token-level early stopping approach that identifies convergence independently at each position. Our method leverages lightweight signals derived from the model's predictions and local context to dynamically determine when individual tokens can be finalized. This yields adaptive per-token freezing without task-specific fine-tuning, substantially reducing the total number of diffusion steps required. Across diverse benchmarks, spanning mathematical reasoning, general question answering, and scientific understanding, our approach achieves state-of-the-art efficiency gains while preserving generation quality. |
| title | Just on Time: Token-Level Early Stopping for Diffusion Language Models |
| topic | Machine Learning Computation and Language |
| url | https://arxiv.org/abs/2602.11133 |