Attend Before Attention: Efficient and Scalable Video Understanding via Autoregressive Gazing
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912963548413952 |
|---|---|
| author | Shi, Baifeng Fu, Stephanie Lian, Long Ye, Hanrong Eigen, David Reite, Aaron Li, Boyi Kautz, Jan Han, Song Chan, David M. Molchanov, Pavlo Darrell, Trevor Yin, Hongxu |
| author_facet | Shi, Baifeng Fu, Stephanie Lian, Long Ye, Hanrong Eigen, David Reite, Aaron Li, Boyi Kautz, Jan Han, Song Chan, David M. Molchanov, Pavlo Darrell, Trevor Yin, Hongxu |
| contents | Multi-modal large language models (MLLMs) have advanced general-purpose video understanding but struggle with long, high-resolution videos -- they process every pixel equally in their vision transformers (ViTs) or LLMs despite significant spatiotemporal redundancy. We introduce AutoGaze, a lightweight module that removes redundant patches before processed by a ViT or an MLLM. Trained with next-token prediction and reinforcement learning, AutoGaze autoregressively selects a minimal set of multi-scale patches that can reconstruct the video within a user-specified error threshold, eliminating redundancy while preserving information. Empirically, AutoGaze reduces visual tokens by 4x-100x and accelerates ViTs and MLLMs by up to 19x, enabling scaling MLLMs to 1K-frame 4K-resolution videos and achieving superior results on video benchmarks (e.g., 67.0% on VideoMME). Furthermore, we introduce HLVid: the first high-resolution, long-form video QA benchmark with 5-minute 4K-resolution videos, where an MLLM scaled with AutoGaze improves over the baseline by 10.1% and outperforms the previous best MLLM by 4.5%. Project page: https://autogaze.github.io/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_12254 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Attend Before Attention: Efficient and Scalable Video Understanding via Autoregressive Gazing Shi, Baifeng Fu, Stephanie Lian, Long Ye, Hanrong Eigen, David Reite, Aaron Li, Boyi Kautz, Jan Han, Song Chan, David M. Molchanov, Pavlo Darrell, Trevor Yin, Hongxu Computer Vision and Pattern Recognition Multi-modal large language models (MLLMs) have advanced general-purpose video understanding but struggle with long, high-resolution videos -- they process every pixel equally in their vision transformers (ViTs) or LLMs despite significant spatiotemporal redundancy. We introduce AutoGaze, a lightweight module that removes redundant patches before processed by a ViT or an MLLM. Trained with next-token prediction and reinforcement learning, AutoGaze autoregressively selects a minimal set of multi-scale patches that can reconstruct the video within a user-specified error threshold, eliminating redundancy while preserving information. Empirically, AutoGaze reduces visual tokens by 4x-100x and accelerates ViTs and MLLMs by up to 19x, enabling scaling MLLMs to 1K-frame 4K-resolution videos and achieving superior results on video benchmarks (e.g., 67.0% on VideoMME). Furthermore, we introduce HLVid: the first high-resolution, long-form video QA benchmark with 5-minute 4K-resolution videos, where an MLLM scaled with AutoGaze improves over the baseline by 10.1% and outperforms the previous best MLLM by 4.5%. Project page: https://autogaze.github.io/. |
| title | Attend Before Attention: Efficient and Scalable Video Understanding via Autoregressive Gazing |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2603.12254 |