新技术在无需重新训练的情况下增强LLM推理能力

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-15 04:00

研究人员推出了一种名为解耦边际锐化（Decoupled Marginal Sharpening, DMS）的新技术，可以在推理时增强大型语言模型（LLM）的推理能力，而无需更新模型权重或使用外部奖励模型。DMS通过分别控制模型对特定推理路径的偏好及其最终答案的选择，解决了现有幂采样方法中的挑战。这种方法允许探索多样化的推理路径，然后更侧重于具有更大聚合支持的答案。在多个基准测试和LLM上的实验表明，DMS在提高推理任务的准确性方面优于现有方法。 AI

影响这种新方法可以在不进行昂贵的重新训练的情况下，提高LLM在复杂任务上的推理性能。

排序理由详细介绍LLM推理时扩展新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 14 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 详细介绍LLM推理时扩展新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: Breaking (< 6h); Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Weixin Wang, Wei Deng, Anderson Schneider, Yuriy Nevmyvaka, Pan Xu, Andrew Bennett · 2026-10-15 04:00

Decoupled Marginal Sharpening for Training-Free Inference-Time Scaling

arXiv:2610.20483v1 Announce Type: new Abstract: We study how to improve LLM reasoning at inference time without updating model weights or using external reward models. Recent reward-free power-sampling methods define the sequence-level power target $\pi_\alpha(y\mid x)\propto p_0…

本文如何被排名

报道来源 [1]

Decoupled Marginal Sharpening for Training-Free Inference-Time Scaling

相关实体

相关话题