📽️📽️ LLaVA-Video (formerly LLaVA-NeXT-Video) has undergone a major upgrade! We are excited to release LLaVA-Video-178K, a high-quality synthetic dataset for video instruction tuning. This dataset includes: - 178,510 caption entries - 960,792 open-ended Q&A pairs -

🙌🏻 New open-sourced dataset and models for video understanding: 📹 LLaVA-Video-178K dataset includes 178K detailed captions and 1.15M Q&A pairs! 📹 LLaVA-Video 7B/72B demonstrate excellent performance on 10+ video benchmarks! 🥰 Truly honored to work with the amazing team!!!

One important goal for video understanding models is ensuring that performance does not degrade as the number of frames increases during inference. We are pleased to see that LLaVA-Video demonstrates this trend in the latest benchmark for long video understanding LongVideoBench.

Learned a lot about autoregressive visual generation and efficient model design from Haoge Deng . Congrats on such a cool project !!!

# LMMs-Eval v0.3.3 Updates 🌟 We added `sglang_srt`, `vllm` and `openai_compatible` models to support wide range opensource and API models and plan to make them as the recommended ways to integrate new models instead of writing different model class. 1. We better support omni

🥰 Thanks AK for sharing our work !!! 🔍 MMSearch-R1 is an end-to-end RL framework that enables LMMs to perform on-demand, multi-turn search with real-world multimodal search tools. 📰 Check out more experimental findings in our paper !!!

🧠Video Thinking Test for Reasoning LLMs🧠 *Video Thinking Test* (📽️Video-TT📽️) is a holistic benchmark to assess the advanced reasoning and understanding correctness/robustness between LLMs and humans #ICCV2025 - Project: zhangyuanhan-ai.github.io/video-tt/ - Data: huggingface.co/datasets/lmms-…

🎯Video Thinking Test (Video-TT) Challenge🎯 We are now hosting the #ICCV2025 🏆Video-TT Challenge🏆. Showcase your best video LLM models and win amazing prizes #ICCV2025 - Challenge: sites.google.com/view/video-tt-… - Registration: codabench.org/competitions/1…

Welcome to checkout our latest work at mini-o3.github.io! 😇

Awesome update! Meet LLaVA-OneVision-1.5!

🫡

Cool Project! Ready to thrive in the Unified Model Era with LMMs-Engine.

💾Agent Personalization in File-System Behavioral Traces💾 #FileGram is a comprehensive framework that grounds agent memory & personalization in file-system behavioral traces. - Project: filegram.choiszt.com - Code: github.com/Synvo-ai/FileG… - Software: habitus.choiszt.com