作者位次 #4/8ACL 2026Poster· Natural Language Processing
Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference
Zhouxuwen · Fangxin Liu · Chao Wang · Xiao Zheng · Hao Zheng · Min He · Li Jiang · Haibing Guan
摘要
Speculative decoding accelerates autoregressive generation by letting draft tokens bypass full verification, but conventional frameworks suffer from frequent false rejections, particularly when draft models produce semantically correct but lexically divergent outputs. In this paper, we present Ca...