1 paper
Haohui Yang, Jiaxing Sun, Xiujun Ma
Power Sampling sharpens a language model's distribution over complete generation trajectories, offering a verifier-free way to improve reasoning at inference time. It also has the…