Mubarak Shah
@ucfmshah
Followers
2K
Following
91
Media
54
Statuses
324
I work in Video Understanding.
Florida, USA
Joined December 2017
🧩 LLMs / Multi-Criteria Decoding. "Inference-Time Alignment of LLMs via User-Specified Multi-Criteria Transfer Decoding". M. F. El Hajj Chehade, S. S. Ghosal, S. Chakraborty, A. Reddy, D. Manocha, H. Zhu, A. S. Bedi.→ New decoding method for aligning LLM outputs with user prefs.
0
0
1
🌐 Knowledge Graph Embeddings / Symbolic Reasoning. "Improving Soft Unification with Knowledge Graph Embedding Methods". Xuanming Cui, Wei Peng Chionh, Adriel Kuek, Ser-Nam Lim.→ Enhancing reasoning over KGs using neural embeddings.
1
0
0
🧬 Dataset Distillation / Diffusion Models. "MGD³: Mode-Guided Dataset Distillation using Diffusion Models". Jeffrey A. Chan-Santiago, Praveen Tirupattur, Gaurav K. Nayak, Gaowen Liu, Mubarak Shah.→ Using diffusion for compact and effective dataset creation. (Oral).
1
0
0
🚫 Zero-Shot Transfer / RL. "Pessimism Principle Can Be Effective: Towards a Framework for Zero-Shot Transfer Reinforcement Learning". Chi Zhang, Ziying Jia, George Atia, Sihong He, Yue Wang.→ Exploring pessimism in RL to enable transfer without fine-tuning.
1
0
0
🎯 Reinforcement Learning / Model Uncertainty. "Efficient and Scalable Reinforcement Learning for Average Reward under Model Uncertainty". Zachary A. Roch, George Atia, Yue Wang.→ Robust RL under uncertainty with improved efficiency.
1
0
0
🔢 Graph Neural Networks / Optimization. "Expressive Power of Graph Neural Networks for (Mixed-Integer) Quadratic Programs". Ziang Chen, Xiaohan Chen, Jialin Liu, Xinshang Wang, Wotao Yin.→ Investigating GNNs' ability to solve complex optimization problems.
1
0
0
📉 Optimization Theory / Matrix Recovery. "Guarantees of a Preconditioned Subgradient Algorithm for Overparameterized Asymmetric Low-Rank Matrix Recovery". Paris Giampouras, HanQin Cai, René Vidal.→ Theoretical guarantees for structured low-rank recovery.
1
0
0
🧠 Embodied Intelligence / Large Models. "LARM: Large Auto-Regressive Model for Long-Horizon Embodied Intelligence". Zhuoling Li, Xiaogang Xu, Zhenhua Xu, Ser-Nam Lim, Hengshuang Zhao.→ Tackling long-horizon tasks with large-scale autoregressive modeling.
1
0
0
📊 Think your VideoQA model reallyunderstands video? .Project Page: Give it a shot on ImplicitQA. We’d love to see your results!. #VideoQA #ComputerVision #MultimodalAI #Research #Dataset #AI #LLM #VideoLLM.
swetha5.github.io
1
0
3
Big shoutout to the CRCV team: @sirnam_swetha, @rohitgUCF and (Parth Parag Kulkarni, David Shatwell, Jeffrey Chan Santiago, Nyle Siddiqui, Joseph Fioresi) for amazing annotation work and Jashua for serving as the human baseline! 🙌.
1
0
3
⚒️ Resources available:.📄 Paper: 📁Dataset: 🧪Eval: 📝Annotation UI: Just plug in your model and see how it holds up.
github.com
Tool for Annotating Videos with Questions and Answers - rohit-gupta/FrameQuiz
1
0
3
🎞️ Diverse Dataset. ImplicitQA spans 15 genres, 7 decades, and media types:. 🟠 58% Animation, 🔵42% Live-action. 🕰️Clips from the 1960s to 2020s. 🎭Genres: Comedy, Action, Adventure, Fantasy, Sci-fi, Drama, Thriller, Romance, and more!.
1
0
2
📊 What’s in the benchmark?.✅ 1,000 expert hand-curated QA pairs. ✅320+ diverse movie clips. ✅9 implicit reasoning types (motion trajectory dynamics, causal, social, spatial…). ✅Human baseline: 83% vs 🤖 Best LLM (GPT-O3): 64%.It’s tough - but that’s the point.
1
0
2
🧠 Why ImplicitQA?.Most VideoQA datasets focus on what’s explicitly shown. But real understanding - like in movies - often requires implicit inference. Check out the sample questions below
1
0
3
🚀New Paper & Dataset out!. ImplicitQA: Going Beyond Frames Toward Implicit Video Reasoning 🎬.We challenge VideoQA models to reason about what’s not visible - like implied actions, hidden causality, and off-screen context. Paper: 🧵👇. #VideoQA #AI #LLM.
arxiv.org
Video QA has made significant strides by leveraging multimodal learning to align visual and textual modalities. However, current benchmarks overwhelmingly focus on questions answerable through...
3
12
47
