| cense: cc-by-nc-nd-4.0 | |
| --- | |
| # PosS: Position Specialist for Speculative Decoding | |
| This repository provides the model checkpoint for **PosS (Position Specialist)**, a speculative decoding method proposed in the paper: | |
| > **PosS: Position Specialist Generates Better Draft for Speculative Decoding** | |
| PosS improves speculative decoding by training a position-specialized draft model that generates higher-quality drafts, leading to improved efficiency and acceptance rates during decoding. | |
| --- | |
| ## 🔗 Code | |
| The full implementation, training details, and evaluation scripts are available at: | |
| 👉 **GitHub:** https://github.com/shrango/PosS | |
| --- | |
| ## 📦 Files | |
| If the model is not automatically downloaded by your framework, you may manually download the following files from this repository: | |
| - `pytorch_model.bin` — model weights | |
| - `config.json` — model configuration | |
| --- | |
| ## 📖 Citation | |
| If you use this model or the PosS method in your research, please cite: | |
| ```bibtex | |
| @misc{huang2025posspositionspecialistgenerates, | |
| title = {POSS: Position Specialist Generates Better Draft for Speculative Decoding}, | |
| author = {Langlin Huang and Chengsong Huang and Jixuan Leng and Di Huang and Jiaxin Huang}, | |
| year = {2025}, | |
| eprint = {2506.03566}, | |
| archivePrefix= {arXiv}, | |
| primaryClass = {cs.CL}, | |
| url = {https://arxiv.org/abs/2506.03566} | |
| } | |