arXiv Computation and Language September 4, 2026 By Prakhar Khatri

Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs

Read the original on arXiv Computation and Language →

The study investigates how long‑video language models decide which frames to keep, compress, and reuse, testing each decision in isolation across six selection rules, three benchmarks, and two answering models. It finds that selecting frames based on queries yields the biggest performance boost, that halving spatial resolution costs little, and that reallocating saved tokens to more compressed frames can further improve accuracy. The work also highlights the importance of a unified evaluation harness to avoid misleading comparisons.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv Computation and Language.

arXiv Computer Vision

Aug 25

RS$^3$-Prune: Read-Sparse, Store-Sparse Token Pruning for Video Object Segmentation

arXiv:2608.22526v1 Announce Type: new Abstract: We introduce RS$^3$-Prune, a training-free token-pruning recipe that instantiates as a small set of inference time hooks atop existing video object seg...

By Avilasha Mandal, Sarvesh Shashikumar

computer-vision efficiency benchmarks

arXiv Computer Vision

Sep 10

MotionBlind: Probing the Illusion of Motion Understanding in Video-LLMs

arXiv:2609.09528v1 Announce Type: new Abstract: Video large language models (Video-LLMs) are increasingly used as the perceptual front end of world models, a role that assumes they can read motion: h...

By Dhairya Bhatia, Bishoy Galoaa, Oliver Fritsche, Shahid Kamal, Muhammad Obaidullah Abdul Salam, Umer Saleem, Om Rastogi, Frania Felix Chettiar, Nesli Erdogmus, Sarah Ostadabbas

llms benchmarks

arXiv AI

Aug 5

When and Where to Look: Adaptive Visual Evidence Scheduling for Efficient Long Video Understanding

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

llms agents multimodal

arXiv AI

Jul 28

Cheap Probes Predict Expensive Training in 3D-CT Vision--Language Models

arXiv:2607. 22771v1 Announce Type: cross Abstract: Picking the frozen image encoder for a 3D~CT vision--language model (VLM), together with the token-compression scheme on top of it, is a search over many candidates.

By Renjie Liang

llms rag fine-tuning multimodal benchmarks

arXiv AI

Aug 13

LoSA: Near-Lossless Sparse Attention for Training-Free Video Diffusion Acceleration

arXiv:2608. 12032v1 Announce Type: cross Abstract: Video diffusion transformers are costly to sample: every denoising step applies self-attention over a long 3D token sequence, a quadratic cost that dominates as resolution and duration grow.

By Enhuai Liu, Yunke Wang, Yutong Wang, Changming Sun, Chang Xu

llms diffusion efficiency

arXiv Machine Learning

Jul 10

Less Data, Faster Convergence: Goal-Driven Data Optimization for Multimodal Instruction Tuning

arXiv:2603. 12478v2 Announce Type: replace-cross Abstract: Multimodal instruction tuning is often compute-inefficient because training budgets are spread across large mixed image-video pools whose utility is highly uneven.

By Rujie Wu, Haozhe Zhao, Hai Ci, Yizhou Wang

fine-tuning multimodal benchmarks

One email a morning, machine-written

Related stories

RS$^3$-Prune: Read-Sparse, Store-Sparse Token Pruning for Video Object Segmentation

MotionBlind: Probing the Illusion of Motion Understanding in Video-LLMs

When and Where to Look: Adaptive Visual Evidence Scheduling for Efficient Long Video Understanding

Cheap Probes Predict Expensive Training in 3D-CT Vision--Language Models

LoSA: Near-Lossless Sparse Attention for Training-Free Video Diffusion Acceleration

Less Data, Faster Convergence: Goal-Driven Data Optimization for Multimodal Instruction Tuning