PulseAugur
中
实时 07:55:51
English(EN) Grounding in the Dark: Text-Only Training for Zero-Shot Video Temporal Localization

AI仅凭文本学习视频定位,弥合模态鸿沟

研究人员开发了一种新颖的方法,仅使用文本描述来训练AI模型理解视频内容,从而在训练期间无需配对的视频-文本数据。该方法通过CLIP或LLM2CLIP等模型的文本嵌入来适配现有的LLM(如Qwen3_8B)。为了在推理时弥合文本和视频之间的模态鸿沟,他们引入了最近实体替换(NER)技术,该技术将视频帧映射到文本嵌入,显著提高了在Charades-STA和QV-Highlights等基准测试上的时间定位性能。 AI

影响 无需昂贵的配对视频-文本数据即可训练视频理解模型,可能减轻标注负担。

排序理由 学术论文,详细介绍了一种新的AI视频时间定位方法。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.AI 阅读 →

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

AI仅凭文本学习视频定位,弥合模态鸿沟

本文如何被排名

Signal score
26 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
学术论文,详细介绍了一种新的AI视频时间定位方法。[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]

  1. arXiv cs.AI TIER_1 English(EN) · Divyanshu Mishra, Sabine Sternig, Rakshith Shetty, Erhan Gundogdu ·

    黑暗中的接地:用于零样本视频时间定位的纯文本训练

    arXiv:2611.06149v1 Announce Type: cross Abstract: Video-LLMs are typically trained on paired video-text data for temporal tasks, yet collecting precise temporal annotations and training on large-scale video datasets are both costly. We investigate whether a text-only LLM, with no…