Improving Diffusion Language Model Decoding through Joint Search in Generation Order and Token Space
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866914306506883072 |
|---|---|
| author | Shen, Yangyi Feng, Tianjian Han, Jiaqi Wang, Wen Chen, Tianlang Shen, Chunhua Leskovec, Jure Ermon, Stefano |
| author_facet | Shen, Yangyi Feng, Tianjian Han, Jiaqi Wang, Wen Chen, Tianlang Shen, Chunhua Leskovec, Jure Ermon, Stefano |
| contents | Diffusion Language Models (DLMs) offer order-agnostic generation that can explore many possible decoding trajectories. However, current decoding methods commit to a single trajectory, limiting exploration in trajectory space. We introduce Order-Token Search to explore this space through jointly searching over generation order and token values. Its core is a likelihood estimator that scores denoising actions, enabling stable pruning and efficient exploration of diverse trajectories. Across mathematical reasoning and coding benchmarks, Order-Token Search consistently outperforms baselines on GSM8K, MATH500, Countdown, and HumanEval (3.1%, 3.8%, 7.9%, and 6.8% absolute over backbone), matching or surpassing diffu-GRPO post-trained d1-LLaDA. Our work establishes joint search as a key component for advancing decoding in DLMs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_20339 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Improving Diffusion Language Model Decoding through Joint Search in Generation Order and Token Space Shen, Yangyi Feng, Tianjian Han, Jiaqi Wang, Wen Chen, Tianlang Shen, Chunhua Leskovec, Jure Ermon, Stefano Computation and Language Machine Learning Diffusion Language Models (DLMs) offer order-agnostic generation that can explore many possible decoding trajectories. However, current decoding methods commit to a single trajectory, limiting exploration in trajectory space. We introduce Order-Token Search to explore this space through jointly searching over generation order and token values. Its core is a likelihood estimator that scores denoising actions, enabling stable pruning and efficient exploration of diverse trajectories. Across mathematical reasoning and coding benchmarks, Order-Token Search consistently outperforms baselines on GSM8K, MATH500, Countdown, and HumanEval (3.1%, 3.8%, 7.9%, and 6.8% absolute over backbone), matching or surpassing diffu-GRPO post-trained d1-LLaDA. Our work establishes joint search as a key component for advancing decoding in DLMs. |
| title | Improving Diffusion Language Model Decoding through Joint Search in Generation Order and Token Space |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2601.20339 |