---
license: mit
language:
- en
tags:
- DLLM
- diffusion-language-model
- on-policy-distillation
- post-training
library_name: transformers
pipeline_tag: text-generation
base_model: Qwen/Qwen3-0.6B
datasets:
- divelab/opdlm_train_data
arxiv: 2606.06712
---
# OPDLM-0.6B

OPDLM-0.6B is a block diffusion language model (DLM) obtained by post-training an
autoregressive language model (ARLM) into a diffusion language model via
**on-policy distillation**. arXiv report: [arxiv.org/abs/2606.06712](https://arxiv.org/abs/2606.06712)

## Highlights
- **Converted, not pretrained from scratch:** built from a strong ARLM, reusing its prior.
- **Training-efficient:** orders of magnitude fewer tokens than from-scratch DLM training (same base ARLM).
- **Inference-efficient:** parallel token decoding via block diffusion.

## Model Details
- **Developed by:** DIVE Lab, Texas A&M University
- **Base model:** [Qwen3-0.6B](https://huggingface.co/Qwen/Qwen3-0.6B)
- **Model type:** Block diffusion language model (decoder-based)
- **Block size:** 4
- **Parameters:** ~0.6B
- **Language:** English
- **License:** MIT

## Training
- **Method:** On-policy distillation from a frozen ARLM teacher into a block DLM student.
- **Conversion budget:** ~<fill in>B tokens
- **Data:** [opdlm_train_data](https://huggingface.co/datasets/divelab/opdlm_train_data)

## Results
For detailed results and benchmarks, please refer to our paper: [arxiv.org/abs/2606.06712](https://arxiv.org/abs/2606.06712)

## Citation
```bibtex
@misc{su2026dataefficientautoregressivetodiffusionlanguagemodels,
      title={Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation}, 
      author={Xingyu Su and Jacob Helwig and Shubham Parashar and Atharv Chagi and Lakshmi Jotsna and Degui Zhi and James Caverlee and Dileep Kalathil and Shuiwang Ji},
      year={2026},
      eprint={2606.06712},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2606.06712},
}
```