flexdraft-flexible-speculative-decoding-via-attention-tuning-and-bonus-guided-calibration
wiki · · 1 min read
FlexDraft: Flexible Speculative Decoding via Attention Tuning and Bonus-Guided Calibration
Definition
arXiv ID: 2605.20022
Sources
- raw/AI/arxiv-cl-260520022-FlexDraft-Flexible-Speculative-Decoding.md