New FACT Method Targets Fixed-Budget Allocation and Cross-Step Training for Speculative Decoding
A new arXiv paper introduces FACT, a method for fixed-budget allocation and cross-step training in speculative decoding, which accelerates language-model generation by verifying multiple draft tokens in one target-model pass.
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-11-19 05:00 · arXiv cs.LG
FACT: Fixed-Budget Allocation and Cross-Step Training for Speculative Decoding