long-video

Tag

Cards List
#long-video

NexSight: Verifiable Tasks for Fine-Grained Agentic Long-Video Understanding

Hugging Face Daily Papers ↗ · 4d ago Cached

NexSight introduces verifiable tasks for fine-grained agentic long-video understanding, presenting SightForge, the 310-task SightLine benchmark, and the SightLoop harness; post-training a 35B model on under 6K verified examples nearly doubles its interactive accuracy and outperforms its 397B family model.

0 favorites 0 likes
#long-video

VideoNexus: Video Perception Beyond the Reasoning Context

Hugging Face Daily Papers ↗ · 5d ago Cached

VideoNexus 是一个免训练框架,将视频细节感知与证据驱动的规划推理分离,由感知子代理返回紧凑的时间锚定证据,在长视频基准上以更低成本取得更高准确率。

0 favorites 0 likes
#long-video

meituan-longcat/LongCat-Video-Avatar-1.6

Hugging Face Models Trending ↗ · 2026-11-13 Cached

美团开源发布了 LongCat-Video-Avatar 1.6,这是其音频驱动人体视频生成框架的最新升级版本,改进了口型同步、长时序视频一致性(可超过10分钟)以及端到端推理速度。

0 favorites 0 likes
#long-video

More Memory, Better Videos? A Controlled Study of Visual Memory in Long-Video Generation

Hugging Face Daily Papers ↗ · 2026-10-26 Cached

本文提出 VMem-Bench,一个两阶段诊断基准,系统研究长视频生成中视觉记忆(检索历史画面)对历史一致性与当前故事准确性的权衡影响,发现检索虽提升实体召回率却降低了精确度与禁用实体规避能力。

0 favorites 0 likes
#long-video

VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token

Hugging Face Daily Papers ↗ · 2026-10-19 Cached

VideoMaMa++ 提出一种掩码引导的扩散式视频抠图框架,通过分离的 RGB/掩码 token 流以及 preserve-and-refine 可学习嵌入,实现跨片段边界的滑窗推理,从而在长视频中保持时间一致性并提升细节精度。

0 favorites 0 likes
#long-video

Elephant in the Fridge: Constant-Memory Frame Packing for Long Video Understanding

Hugging Face Daily Papers ↗ · 2026-10-12 Cached

提出 FramePack-U,一种免训练的常量内存帧打包框架,可将任意长度视频压缩为固定大小的视觉表示,直接插入现成 VLM 推理,在 Long VideoBench、MLVU 和 Video-MME 上显著超越全分辨率基线及现有 token 压缩方法。

0 favorites 0 likes
#long-video

@denziideng: Meituan has jumped into AI video production, and with their first move, they've dropped a 13.6B model—long videos with …

X AI KOLs Timeline ↗ · 2026-10-06 Cached

美团开源了13.6B参数的LongCat-Video视频生成模型,原生支持文本到视频、图像到视频和视频续写,可在720p/30fps下生成数分钟长视频且无色彩衰减或质量下降,权重采用MIT许可证并允许商业使用。

0 favorites 0 likes
#long-video

In-Distribution Forcing for Long Video Generation at Test Time

Hugging Face Daily Papers ↗ · 2026-10-02 Cached

提出 In-Distribution Forcing (ID-Forcing) 测试时框架,通过自缓存机制避免 KV 条目偏离训练分布,从而将短时视频扩散模型扩展到分钟级长视频生成,并显著缓解漂移问题。

0 favorites 0 likes
#long-video

CoEvoWhen: Policy-Tool Coevolution for Ultra-Long Video Temporal Grounding

Hugging Face Daily Papers ↗ · 2026-09-30 Cached

CoEvoWhen proposes a policy-tool coevolution framework that jointly evolves high-level policies and executable media tools from VLM agentic reasoning trajectories to improve ultra-long video temporal grounding without updating model parameters.

0 favorites 0 likes
#long-video

KeyRec: Bounded Visual Memory for Streaming and Long-Video Understanding

Hugging Face Daily Papers ↗ · 2026-09-26 Cached

KeyRec is a training-free framework for building bounded visual memory in vision-language models, using a recent-observation cache plus a structured event bank to compress long/streaming video context to just 10% of dense visual tokens while setting state-of-the-art results on 13 of 15 streaming and long-video benchmark settings.

0 favorites 0 likes
#long-video

Open Omnimodal World Models (GitHub Repo)

TLDR AI ↗ · 2026-08-26 Cached

JoyAI-Echo is an open-source GitHub repository that provides long-horizon audio-visual generation and omnimodal world models for creating persistent stories and interactive worlds.

0 favorites 0 likes
#long-video

NARU: A Benchmark for NARrative Evolution and Cultural Nuance Understanding in Japanese Extreme Long Video

Hugging Face Daily Papers ↗ · 2026-08-13 Cached

NARU is a benchmark for evaluating narrative evolution and cultural reasoning in Japanese extreme long-form videos, constructed through a hierarchical annotation pipeline and native-speaker verification to address gaps in current benchmarks.

0 favorites 0 likes
#long-video

SCOUT: Self-Checking and Recovery-Aware Tool-Thought Agents for Ultra-Long Egocentric Video Reasoning

arXiv cs.AI ↗ · 2026-08-11 Cached

SCOUT introduces a recovery-aware agentic framework with an adaptive exploration-exploitation policy for ultra-long egocentric video reasoning, trained via UPS-GRPO, a uncertainty-prioritized RL method with turn-level advantage decomposition. It achieves state-of-the-art results on ultra-long egocentric benchmarks and remains competitive on shorter long-video settings.

0 favorites 0 likes
#long-video

Self Gradient Forcing: Native Long Video Extrapolation

Papers with Code Trending ↗ · 2026-07-22 Cached

Proposes Self Gradient Forcing (SGF), a two-pass training strategy for autoregressive video diffusion models that provides missing supervision for writing useful context memory, enabling strong long-video extrapolation even from short training windows.

0 favorites 0 likes
#long-video

@svpino: This model can generate coherent 1+ hour videos across multiple scenes without skipping a beat. I read their paper so y…

X AI KOLs Following ↗ · 2026-07-13 Cached

This tweet explains LingBot-World-Infinity, an open-weight video generation model that uses a training technique to recover from errors, enabling coherent hour-long videos across multiple scenes.

0 favorites 0 likes
#long-video

OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators

Hugging Face Daily Papers ↗ · 2026-07-09 Cached

OPSD-V improves few-step autoregressive video diffusion models by using real long-video data as temporal context during training, providing dense trajectory-level supervision that enhances visual quality and motion dynamics without altering inference mechanisms.

0 favorites 0 likes
#long-video

@IndieDevHailey: HKU open-sources ViMax — AI one-person film crew, one-click to produce a complete long-form video! One-sentence idea → automatically generates a finished production with script, storyboard, voiceover, and consistent characters! Solves these pain points: - No more short clips of a few seconds where characters change faces - Multi-agent collaboration: screenwriter writes story, director controls shots, producer ensures consistency, generator outputs video...

X AI KOLs Timeline ↗ · 2026-07-06 Cached

HKU open-sourced ViMax, a multi-agent collaborative long video generation tool that can generate a coherent video with script, storyboard, voiceover, and consistent characters from a single sentence, solving issues like short video fragmentation and character inconsistency. The developer also introduced Taste-Skill, a frontend framework that improves the aesthetics of AI-generated interfaces.

0 favorites 0 likes
#long-video

@VraserX: Seedance 2.5 making consistent videos up to 3 minutes in one go is honestly insane. If this pace continues, we’ll be ma…

X AI KOLs Following ↗ · 2026-07-03 Cached

Seedance 2.5 enables consistent AI-generated videos up to 3 minutes in one go, marking a significant leap in AI video generation that could soon lead to TV episodes and feature films.

0 favorites 0 likes
#long-video

Rethinking RAG in Long Videos: What to Retrieve and How to Use It?

arXiv cs.AI ↗ · 2026-06-12 Cached

This paper introduces V-RAGBench, a benchmark for evaluating retrieval-augmented generation over long egocentric videos, and CARVE, a method that adaptively selects retrieval configurations per chunk to improve VideoRAG performance.

0 favorites 0 likes
#long-video

OmniMem: Perturbation-aware Memory Compression for Streaming Audio-Visual LLMs

arXiv cs.AI ↗ · 2026-06-09 Cached

OmniMem introduces a modality-aware memory allocation and perturbation-aware selection strategy for streaming audio-visual LLMs, achieving 2-4% absolute accuracy gains over compression baselines on long-video benchmarks.

0 favorites 0 likes
Next →
← Back to home

Submit Feedback