WhiFlash-Neural Collection Cross-paradigm speculative decoding method to accelerate LLM inference (Kwon et al., Findings of EMNLP 2026). • 7 items • Updated 19 days ago