# Long-form video understanding

3 related articles

Subscribe to updates

arXiv Oct 6, 2026
FlashGaze prunes video patches before ViT encoding with Quadtree DP, holding 98% of LongVideoBench accuracy at a 20% retained ratio on Qwen3-VL-8B while cutting peak GPU memory by 1.8x
FlashGaze is a training-free method that reduces spatiotemporal redundancy before ViT encoding by using pixel-space differences as a proxy for information loss and Quadtree Dynamic Programming to jointly decide patch dropping, merging, and keeping under a fixed budget; on Qwen3-VL-8B and NVILA-8B-Video it achieves higher accuracy and lower TTFT across multiple video understanding benchmarks, and at a 20% retained ratio preserves 98% of the full-input LongVideoBench accuracy.
FlashGaze is a training-free method that reduces spatiotemporal redundancy before ViT encoding by using pixel-space differences as a proxy for information loss and Quadtree Dynamic Programming to jointly decide patch dropping, merging, and keeping under a fixed budget; on Qwen3-VL-8B and NVILA-8B-Video it achieves higher accuracy and lower TTFT across multiple video understanding benchmarks, and at a 20% retained ratio preserves 98% of the full-input LongVideoBench accuracy.
Ziye Zhu（First author）
FlashGaze Visual Token Pruning Quadtree Dynamic Programming Long-form video understanding Multimodal Large Language Models
arXiv Oct 1, 2026
LEAP keeps hour-scale audio-video QA out of one giant context: 4.5–16.8% over the Qwen3-Omni-30B baseline and transfer to MiniCPM-o 4.5
LEAP splits an hour-scale recording into fixed-duration blocks, runs a lightweight localization pass per block to score short candidate windows, then re-encodes only the top-ranked windows in a single bounded answer pass, so the answer input and peak context stay independent of recording duration; across several AVQA benchmarks it improves over the Qwen3-Omni-30B-A3B baseline by 4.5–16.8% and transfers to MiniCPM-o 4.5, surpassing its published results by 3.1–13.0%.
LEAP splits an hour-scale recording into fixed-duration blocks, runs a lightweight localization pass per block to score short candidate windows, then re-encodes only the top-ranked windows in a single bounded answer pass, so the answer input and peak context stay independent of recording duration; across several AVQA benchmarks it improves over the Qwen3-Omni-30B-A3B baseline by 4.5–16.8% and transfers to MiniCPM-o 4.5, surpassing its published results by 3.1–13.0%.
Futurewei Technologies (United States)Northeastern University
Juyi Lin（First author）
LEAP Block-wise Evidence Retrieval Audio-Visual Question Answering Long-form video understanding
arXiv Sep 30, 2026
VideoLoop uses dual-loop bounded working memory to ease semantic thrashing in long-video agents, lifting Gemini 3.1 Pro by 3.2 to 4.5 points on VideoMME (long) and two other benchmarks
The work formalizes a failure mode it calls semantic thrashing, in which append-only working memory keeps accumulating noise and dilutes key evidence, and proposes VideoLoop: an outer multimodal agent explores the video inside a sandboxed filesystem while an inner memory orchestrator retrieves relevant artifacts from an unbounded filesystem and rewrites a bounded working memory after every step; across VideoMME (long), VideoMMMU, and LongVideoBench (long), VideoLoop improves four LVLM backbones in a plug-and-play manner, averaging a 4.2-point gain over baseline on VideoMME (long) and reaching 88.3%, 88.8%, and 80.9% with Gemini 3.1 Pro.
The work formalizes a failure mode it calls semantic thrashing, in which append-only working memory keeps accumulating noise and dilutes key evidence, and proposes VideoLoop: an outer multimodal agent explores the video inside a sandboxed filesystem while an inner memory orchestrator retrieves relevant artifacts from an unbounded filesystem and rewrites a bounded working memory after every step; across VideoMME (long), VideoMMMU, and LongVideoBench (long), VideoLoop improves four LVLM backbones in a plug-and-play manner, averaging a 4.2-point gain over baseline on VideoMME (long) and reaching 88.3%, 88.8%, and 80.9% with Gemini 3.1 Pro.
Microsoft Microsoft (United States)University of Rochester
Jianming Xu（First author）
VideoLoop Dual-loop bounded working memory Semantic thrashing Long-form video understanding Agent Memory

You’re all caught up

FlashGaze prunes video patches before ViT encoding with Quadtree DP, holding 98% of LongVideoBench accuracy at a 20% retained ratio on Qwen3-VL-8B while cutting peak GPU memory by 1.8x

LEAP keeps hour-scale audio-video QA out of one giant context: 4.5–16.8% over the Qwen3-Omni-30B baseline and transfer to MiniCPM-o 4.5

VideoLoop uses dual-loop bounded working memory to ease semantic thrashing in long-video agents, lifting Gemini 3.1 Pro by 3.2 to 4.5 points on VideoMME (long) and two other benchmarks