#MultiModalLLM - Twitter Hashtag

13 days ago

📢 We are pleased to invite submissions to the 2nd MUCG Workshop @ ECCV 2026! 🌐 https://t.co/klg6BbpWUz 📍 Malmö, Swede 📝 Submission Deadline: July 1, 2026 📚 Archival & Non-Archival Tracks 🏆 Best Paper Awards #ECCV2026 #ECCV #MultimodalLLM #GenerativeAI #MUCG

SQWu_Tori's tweet photo. 📢 We are pleased to invite submissions to the 2nd MUCG Workshop @ ECCV 2026!

🌐 https://t.co/klg6BbpWUz
📍 Malmö, Swede
📝 Submission Deadline: July 1, 2026
📚 Archival & Non-Archival Tracks
🏆 Best Paper Awards
#ECCV2026 #ECCV #MultimodalLLM #GenerativeAI #MUCG https://t.co/aZq5q2dMmh

0

15

7

1

6K

Jiaqi Tang @jqtust

16 days ago

Happy to introduce Robust-U1 #ICML — a framework that teaches MLLMs to self-recover corrupted visual content for robust understanding. 📄 Paper: https://t.co/qEg4uKQagq 💻 Code: https://t.co/mzbrhTruI2 🌍 Demo: https://t.co/biSy92E39i #RobustAI #MultimodalLLM #ComputerVision

0

2

1

0

106

Zhuokai Zhao

@zhuokaiz

about 1 year ago

Happy to share that DYTO has been accepted to #ICCV2025 as a new SOTA in video understanding! Try it out in your experiments and see how it measures up! Paper: https://t.co/ATvvSiGcWn Code: https://t.co/SINMWnEdG5 #MultimodalAI #MultimodalLLM #LargeLanguageModel #LLM #VisionLanguageModel #VLM #LVLM #LargeVisionLanguageModel #AIReasoning #ComputerVision #MachineLearning #DeepLearning #AIResearch #ICCV2025 #ICCV #CVPR

Zhuokai Zhao

@zhuokaiz

over 1 year ago

We've ( @Yiming1254115 @ZRChen_AISafety ) been wondering why so many existing video understanding approaches stick to uniformly sampling a fixed number of frames (e.g., 8) from an entire video sequence as input. So...... Introducing DYTO, a training-free framework that dynamically balances token efficiency and semantic richness. DYTO sets state-of-the-art performance across benchmarks with zero-shot adaptability. 🎥💡 Key Highlights: 🔹Hierarchical Frame Selection: Dynamically captures crucial events without rigid sampling. 🔹Bipartite Token Merging: Reduces token redundancy while preserving rich contextual details. 🔹Best-in-class Performance: Outperforms both fine-tuned & training-free methods. Read more: https://t.co/gIkTzEpXfl #AI #ML #VideoUnderstanding #LLM #MLLM #LVLM #ZeroShot #MultiModal

zhuokaiz's tweet photo. We've ( @Yiming1254115 @ZRChen_AISafety ) been wondering why so many existing video understanding approaches stick to uniformly sampling a fixed number of frames (e.g., 8) from an entire video sequence as input. So......

Introducing DYTO, a training-free framework that dynamically balances token efficiency and semantic richness. DYTO sets state-of-the-art performance across benchmarks with zero-shot adaptability. 🎥💡

Key Highlights:
🔹Hierarchical Frame Selection: Dynamically captures crucial events without rigid sampling.
🔹Bipartite Token Merging: Reduces token redundancy while preserving rich contextual details.
🔹Best-in-class Performance: Outperforms both fine-tuned & training-free methods.

Read more: https://t.co/gIkTzEpXfl

#AI #ML #VideoUnderstanding #LLM #MLLM #LVLM #ZeroShot #MultiModal

1

5

4

2K

0

12

1

0

478

Yinghao Ma @nicolaus625

about 1 year ago

CMI-Bench bridges NLP and music information retrieval. Let’s build smarter, fairer, and more musical LLMs! #CMIBench #MusicAI #MultimodalLLM #AudioLLM #MIR #ISMIR2025 #Huggingface

0

2

0

136

NEUZEN AI @Neuzenai23

about 1 year ago

🚨Meta launches Llama 4 Scout & Maverick — the most powerful multimodal LLMs to date. 📍Scout: 10M token context, 16 experts 📍Maverick: 128 experts, best image grounding 🔍 More info: https://t.co/h8Hecea2yu #Llama4 #MetaAI #AI #MultimodalLLM #GenAI #TechNews