ShallowStream is a framework for streaming video understanding that uses the shallow layers of a multimodal large language model (MLLM) to encode frames and build a lightweight index. During streaming, it maintains an always‑on index via the KV cache of shallow layers, and at query time it scores context frames using shallow‑layer attention and selects diverse evidence for answering. The approach matches the performance of leading streaming methods while cutting per‑frame prefill latency and 10‑second end‑to‑end latency by up to 52.1× and 11.9×, respectively.
arXiv:2609.16722v1 Announce Type: new Abstract: Scaling Multimodal Large Language Models (MLLMs) to long-form video understanding is bottlenecked by the explosion of visual tokens, which saturates co...
Scaling Multimodal Large Language Models (MLLMs) to long-form video understanding is bottlenecked by the explosion of visual tokens, which saturates context windows and incurs prohibitive costs. Curre...
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.
arXiv:2510. 09608v2 Announce Type: replace-cross Abstract: Vision-language models (VLMs) could power real-time assistants and autonomous agents, but they face a critical challenge: understanding near-infinite video streams without escalating latency and memory usage.
arXiv:2607. 24794v1 Announce Type: new Abstract: While Multimodal Large Language Models (MLLMs) demonstrate superior generalization in fundamental video tasks, restricted context windows limit their long video understanding.
The paper introduces MaLoW, a streaming memory framework that internalizes long‑term video history into evolving lightweight parameter updates using a recurrent HyperNetwork and LoRA weights. MaLoW keeps current observations as input while carrying historical context forward without re‑visiting the full video stream. The authors also present MaLoW‑Instruct, a large streaming video dataset, and show that MaLoW outperforms prior memory methods on several benchmarks and improves action success in a vision‑language‑action setting.
arXiv:2607. 25266v2 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) have enabled long-form video understanding at a scale that was not previously possible.
arXiv:2504. 01407v3 Announce Type: replace-cross Abstract: Long video understanding poses a fundamental challenge for large video-language models (LVLMs) due to the overwhelming number of frames and the risk of losing essential context through naive downsampling.
arXiv:2608.20473v1 Announce Type: new Abstract: Video language models process videos as dense visual-token sequences with substantial representational redundancy. Compressing these sequences is there...
arXiv:2604.17422v2 Announce Type: replace Abstract: Long video understanding remains a formidable challenge for Multimodal Large Language Models (MLLMs) due to the prohibitive cost of processing dens...
arXiv:2605.08158v2 Announce Type: replace-cross Abstract: Long video understanding with multimodal language models suffers from three compounding bottlenecks: heavy decode cost to obtain dense RGB fr...
FastBench is a new benchmark that tests streaming Video‑Large‑Language Models (VLMs) on high‑dynamic real‑world video streams, where events occur rapidly and cannot be captured by sparse 1–2 FPS sampling. The benchmark uses a trajectory‑grounded pipeline that generates QA pairs from high‑FPS clips, filters questions answerable at 2 FPS, verifies answers with SAM3 and CoTracker3 trajectories, and incorporates three rounds of human inspection, yielding 306 QA pairs across eight domains and multiple temporal scopes. Experiments show that even the strongest model, Gemini‑3.5‑Flash, achieves only 50.7 % accuracy, and that denser sampling improves performance but saturates when history is compressed; a training‑free baseline, ProactiveFrame, outperforms uniform sampling but still falls short of oracle‑guided focusing.