研究人员开发了辅助粒子功率采样（APPS），这是一种新颖的块状粒子算法，旨在提高大型语言模型推理的效率。APPS 旨在更好地定位基础 LLM 已分配概率质量但难以找到的正确多步解决方案。通过在竞争性前缀之间重新分配计算并使用未来价值引导的选择，APPS 增强了推理基准上无训练解码的准确性-运行时权衡。 AI
影响 提高了 LLM 在复杂推理任务中推理的效率，有可能缩小与后训练系统的差距。
排序理由 这是一篇详细介绍 LLM 推理新算法的研究论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2605.02427v1 Announce Type: cross Abstract: A recurring pattern in "reasoning without training" is that base LLMs already assign non-trivial probability mass to correct multi-step solutions; the bottleneck is locating these modes efficiently at inference time. Power samplin…
A recurring pattern in "reasoning without training" is that base LLMs already assign non-trivial probability mass to correct multi-step solutions; the bottleneck is locating these modes efficiently at inference time. Power sampling provides a principled way to bias decoding towar…