English(EN) Test-Time Training with Next-Token Prediction

新的TTT-NTP方法利用下一个词元预测提升LLM性能

作者 PulseAugur 编辑部 · [1 个来源] · 2026-09-01 04:00

研究人员推出了一种名为“测试时训练与下一个词元预测”（TTT-NTP）的新方法，该方法可增强预训练长上下文语言模型的性能。该技术利用语言模型训练中使用的固有下一个词元预测信号，来指导测试时训练过程中的适应过程。与使用代理的先前方法不同，TTT-NTP直接使用模型自身的上下文隐藏状态来监督更新，确保每个适应步骤都与核心训练目标保持一致。在RULER Full-13和LongBench-v2等基准测试上的评估显示，包括Llama-3.1-8B和Mistral-7B-v0.3在内的各种模型在不损害现有能力的情况下，性能得到了一致提升。 AI

影响该方法可以通过利用已部署LLM的核心训练目标来提高其适应性和性能。

排序理由该集群包含一篇详细介绍语言模型适应新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 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: 24 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍语言模型适应新方法的学术论文。[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.CL TIER_1 English(EN) · Xuan Ouyang, Zefan Cai, Junjie Hu · 2026-09-01 04:00

Test-Time Training with Next-Token Prediction

arXiv:2606.21803v2 Announce Type: replace Abstract: Next-token prediction is the self-supervised signal that trains language models, and every observed prompt token provides the same signal at test time. We study whether this signal can define the inner-loop objective for test-ti…

本文如何被排名

报道来源 [1]

Test-Time Training with Next-Token Prediction

相关实体

相关话题