HuggingFace Daily Papers（社区热门论文）· HuggingFace Daily Papers（社区热门论文）·· 5 天前AI 评分44
Before It Fades：在推理阶段强化 VideoLLM 的时间表征
AI 导读
研究提出 Temporal Activation Injection（TAI），一种无需训练的方法，通过在推理阶段将中间层的时间散度向量 τ_l 按衰减规律重新注入后续层，解决 VideoLLM 时间推理能力随层数加深而衰减的问题。该方法在三个 VideoLLM 和四个 benchmark 上一致提升时间推理表现，对非时间任务影响可忽略。
正文
Abstract:Video Large Language Models (VideoLLMs) receive frames in sequential order and interpret how visual content evolves along the temporal axis, yet temporal reasoning remains a persistent weakness across architectures. Reversing the frame order of a video, a transformation that should invert temporal answers, often leaves the final prediction unchanged. We investigate where this failure originates by defining the temporal divergence vector $\tau_l$, the layer-wise representational difference induced by reversing temporal order. Tracking its magnitude across layers reveals a consistent temporal divergence profile where the divergence peaks at intermediate layers and progressively diminishes toward the output. We confirm this peak is specific to temporal reasoning and functionally critical for predictions, establishing that VideoLLMs acquire temporal information at intermediate layers but fail to maintain it to the output. This progressive fading motivates our method, Temporal Activation Injection (TAI), which extracts $\tau_l$ at the peak of the profile for each input and reinjects it into subsequent layers following the measured decay. TAI requires no training and consistently improves temporal reasoning across three VideoLLMs and four benchmarks with negligible impact on non-temporal tasks. Code is available at this https URL.
| Comments: | Accepted to NeurIPS 2026 |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.01595 [cs.CV] |
| (or arXiv:2610.01595v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.01595 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Youngwoo Shin [view email]
[v1]
Thu, 1 Oct 2026 12:45:01 UTC (5,879 KB)
来源：HuggingFace Daily Papers（社区热门论文） · arxiv.org