The paper introduces an audio-first triage method for budgeted vision‑language captioning of untrimmed egocentric video. By selecting windows for a vision‑language model based on lightweight audio features before any video decoding, the approach reduces costly model calls. It achieves 4.0–10.8 percentage point improvements in action coverage across call rates, cuts 9–20% of VLM calls at matched coverage on EPIC‑KITCHENS‑100, and outperforms uniform sampling and recent visual keyframe selectors on Ego4D.
By Masoud Jalayer, Changyi Li, Yu Xiao
LEAP is a framework for long audio‑video question answering that avoids encoding entire recordings by dividing them into fixed‑duration blocks. It performs a lightweight localization pass on each block to score short candidate windows, then pools the highest‑ranked windows for a single bounded answer pass, keeping the answer input and peak context independent of recording length. The method trains both a localization LoRA and an answer LoRA, supports causal streaming queries, and achieves significant performance gains over baseline models on multiple AVQA benchmarks.
By Juyi Lin, Zhiqiang Lao, Jiali Cui, Lin Zhao, Pu Zhao, Dichang Zhang, Arman Akbari, Yu Qi, Xinru Jiang, Yanzhi Wang, Heather Yu, Liang Peng
The paper introduces Cue2Narrate, a two‑stage pipeline that jointly predicts what visual events to narrate and when to insert the narration in long, untrimmed movie clips. It uses a dual‑head audio‑visual localizer to identify visual cue and narration windows, followed by a LoRA‑adapted vision‑language model that generates concise audio descriptions, trained with a Description Ranking Loss. The authors also present the LongLSMDC benchmark, comprising up to 8‑minute clips, and show that Cue2Narrate outperforms video‑only and audio‑only baselines by 5–12 points in average mAP and improves AD generation over fine‑tuned base VLMs.
By Akshita Gupta, Aditya Arora, Federico Tombari, Marcus Rohrbach, Anna Rohrbach
Video-HolmesV2 is a new benchmark that tests multimodal large language models on their ability to reason with spatio‑temporal audio‑visual evidence in long videos. It requires models to justify answers with precise evidence, uses a multi‑model cross‑verification pipeline and a spatio‑temporal evidence‑aware metric, and introduces an audio‑text guided token compression framework to reduce long‑context noise. In evaluations, even strong proprietary models score below 60% while the proposed approach outperforms comparable open‑source omni‑models.
By Zhaoyang Wei, Zipeng Wang, Yushe Cao, Chenhui Qiang, Shuaibing Cheng, Xuesong Yang, Sen Nie, Bowen Jiang, Wenchao Ding, Yanchao Hao, Zheng Wei, Xuehui Yu, Zhenjun Han
CoFiE introduces a two‑stage evidence selection framework for streaming video understanding, separating a coarse, query‑agnostic filtering of visually distinctive frames from a fine, query‑specific refinement during LLM prefill. By filtering out redundant frames before expensive vision encoding, CoFiE reduces end‑to‑end latency while maintaining high accuracy. The method achieves state‑of‑the‑art performance on benchmarks such as StreamingBench and OvO‑Bench, improving accuracy by up to 3.15% and inference speed by up to 2.54× compared to prior approaches.
By Jing Jiang, Yiran Ling, Ruonan Li, Dimitrios Stamoulis, Jie Liu
arXiv:2608. 03918v1 Announce Type: cross Abstract: Efficient long-video understanding requires vision--language models (VLMs) to reason over a small number of frames selected as sparse visual evidence.
By Ke Li, Jiayu Chen, Maoliang Li, Zihao Zheng, Hailong Zou, Hengyi Zhang, Xuanzhe Liu, Xiang Chen
arXiv:2610.08192v1 Announce Type: new
Abstract: Audio-visual models improve egocentric action recognition by exploiting complementary cues, yet typically assume that both streams remain available at...
By Souptik Sen, Zahra Ahmadi
The paper introduces TRACE, a training‑free agent for long‑video understanding that grounds answers in raw visual clips and builds an evidence bundle until the answer stabilises. It also presents VES‑Bench, a 600‑question benchmark over 348 public long videos that audits whether decoded frames cover all necessary evidence intervals at three strictness levels. TRACE achieves high accuracy on VES‑Bench (63.5% audit accuracy) and remains competitive on other video‑understanding benchmarks while using far fewer frames than uniform decoding.
By Pengyiang Liu, Junbo Niu, Xiaoyang Hu, Zhongyue Shi, Zitian Wang, Linjiang Huang, Si Liu
Multimodal Large Language Models have demonstrated impressive video understanding, yet their ability to reason over long-form narratives is often masked by visual-centric evaluations and inefficient c...
MoG-VLM is a motion‑guided vision‑language model designed for efficient video anomaly detection. It uses a lightweight motion signal, Temporal Gradient Magnitude, to perform content‑aware sampling and activates dense multi‑frame VLM reasoning only around frames with high anomaly likelihood via a hierarchical dual‑stream design. The method achieves state‑of‑the‑art AUC‑ROC scores on four benchmarks and delivers a 1.31× inference speedup on XD‑Violence.
By Khaled Waleed Dawoud, Muhammad Zaigham Zaheer, Khan Muhammad, Muhammad Haris Khan
The paper introduces OTFS, a principled optimal transport framework for selecting frames in long-video understanding. It formulates frame selection as a structured evidence-allocation problem, using a semi‑unbalanced entropic optimal transport objective and an efficient OTFS‑Sinkhorn solver, followed by a dynamic‑programming extraction step. Experiments on three benchmarks show OTFS consistently outperforms strong training‑free baselines.
By Ziwei Zhao, Yuhang Liu
MiniVer‑V is a new benchmark for short‑video fact‑checking that contains 195 videos, each annotated with a supported, refuted, or insufficient verdict and 5,510 multimodal evidence units (visual keyframes, speech transcripts, and web‑retrieved sources). The authors propose a two‑layer verification framework that first checks claim‑video consistency using internal evidence, then determines the factual verdict with external corroboration, and employs a sufficiency‑driven greedy search to assemble only the minimal evidence needed for a confident decision. Using Claude Sonnet 4, the method achieves a Macro‑F1 of 0.510 with an average of 4.5 evidence units—about 16 % of the full set—while matching the full‑evidence baseline and improving detection of insufficient cases.
By Leran Chen, Lingnan Kong, Zile Cai