Power-SMC：免训练序列级幂采样加速LLM推理

1. 12:00
   arXiv · Statistics Machine Learning

   Power-SMC：面向免训练 LLM 推理的低延迟序列级幂采样

   研究者提出 Power-SMC，一种免训练采样方法，面向与 Metropolis–Hastings 采样相同的序列级幂分布，同时保持接近标准解码的延迟。该方法并行维护多条候选序列，用重要性权重打分并周期性剪枝，理论证明采样温度 τ=1/α 可唯一消除逐步权重方差。

Current heat 6·Comparable peak 10（Oct 5）·Comparable change over 24 hours –

The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.