1
Variational Speculative Decoding: Rethinking Draft Training from Token Likelihood to Sequence Acceptance
推出变分推测解码,从单路径优化转向序列接受,革新大模型推理加速
arXiv:2602.05774v4 Announce Type: replace Abstract: Speculative decoding accelerates inference for (M)LLMs, yet a training-decoding discrepancy persis…