power-sampling

PreSTO 提出一种预测性子树预取方法加速 PowerMH 功率采样器，通过批量预取未来的 MH 提案并保持转移核不变，在七个数据集和五个 LLM 上实现 1.71×–2.33× 的中位加速。

Chopthin-Consensus Power Sampling (CCPS) is a diversity-preserving method for LLM decoding that improves reasoning accuracy without post-training by preserving distinct reasoning paths and using semantic-majority selection, as demonstrated in benchmarks.

This paper introduces Entropy-Guided Power Sampling (EGPS), a training-free and verifier-free sampler that improves the efficiency of power sampling for enhancing base language model reasoning. EGPS achieves up to 12.6x speedup over standard Metropolis-Hastings sampling while reaching best or tied-best accuracy on benchmarks like MATH500, HumanEval, and GPQA.