研究人员开发了一个名为局部分支路由（LBR）的新框架，以在测试时扩展期间提高语言模型的推理能力。LBR 在 token 级别运行，扩展局部前瞻树并使用轻量级路由器选择最有希望的分支。该方法通过利用候选未来的隐藏状态，实现了更高效且可训练的扩展，在数学推理基准测试中通过提高 Pass@1 和 Pass@32 分数而优于现有方法。 AI
影响 这项研究引入了一种更有效且可训练的改进语言模型推理的方法，有望在复杂任务上取得更好的性能。
排序理由 该集群描述了一篇详细介绍语言模型测试时扩展新颖方法的最新研究论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2606.25354v1 Announce Type: new Abstract: Test-time scaling improves language-model reasoning, but existing approaches often face a difficult trade-off: long chain-of-thought sampling remains single-threaded, while sentence- or solution-level search can be computationally e…
Test-time scaling improves language-model reasoning, but existing approaches often face a difficult trade-off: long chain-of-thought sampling remains single-threaded, while sentence- or solution-level search can be computationally expensive and hard to train end-to-end. We introd…
Test-time scaling improves language-model reasoning, but existing approaches often face a difficult trade-off: long chain-of-thought sampling remains single-threaded, while sentence- or solution-level search can be computationally expensive and hard to train end-to-end. We introd…