AdaSD: Adaptive Speculative Decoding for Efficient Language Model Inference
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866916046107049984 |
|---|---|
| author | Lu, Kuan-Wei Hong, Ding-Yong Liu, Pangfeng Wu, Jan-Jan |
| author_facet | Lu, Kuan-Wei Hong, Ding-Yong Liu, Pangfeng Wu, Jan-Jan |
| contents | Large language models (LLMs) have achieved remarkable performance across a wide range of tasks, but their increasing parameter sizes significantly slow down inference. Speculative decoding mitigates this issue by leveraging a smaller draft model to predict candidate tokens, which are then verified by a larger target model. However, existing approaches often require additional training, extensive hyperparameter tuning, or prior analysis of models and tasks before deployment. In this paper, we propose Adaptive Speculative Decoding (AdaSD), a hyperparameter-free decoding scheme that dynamically adjusts generation length and acceptance criteria during inference. AdaSD introduces two adaptive components: one to determine when to stop candidate token generation and the other to decide token acceptance, updated in real time based on token entropy and Jensen-Shannon distance. This approach eliminates the need for pre-analysis or fine-tuning and is compatible with off-the-shelf models. Experiments on benchmark datasets demonstrate that AdaSD achieves up to 1.46x speedup over vanilla speculative decoding while limiting accuracy degradation to under 1.8%, making it a practical solution for efficient and adaptive LLM inference. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_11280 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AdaSD: Adaptive Speculative Decoding for Efficient Language Model Inference Lu, Kuan-Wei Hong, Ding-Yong Liu, Pangfeng Wu, Jan-Jan Computation and Language Large language models (LLMs) have achieved remarkable performance across a wide range of tasks, but their increasing parameter sizes significantly slow down inference. Speculative decoding mitigates this issue by leveraging a smaller draft model to predict candidate tokens, which are then verified by a larger target model. However, existing approaches often require additional training, extensive hyperparameter tuning, or prior analysis of models and tasks before deployment. In this paper, we propose Adaptive Speculative Decoding (AdaSD), a hyperparameter-free decoding scheme that dynamically adjusts generation length and acceptance criteria during inference. AdaSD introduces two adaptive components: one to determine when to stop candidate token generation and the other to decide token acceptance, updated in real time based on token entropy and Jensen-Shannon distance. This approach eliminates the need for pre-analysis or fine-tuning and is compatible with off-the-shelf models. Experiments on benchmark datasets demonstrate that AdaSD achieves up to 1.46x speedup over vanilla speculative decoding while limiting accuracy degradation to under 1.8%, making it a practical solution for efficient and adaptive LLM inference. |
| title | AdaSD: Adaptive Speculative Decoding for Efficient Language Model Inference |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2512.11280 |