Daily Papers of 2025-11-20

1. Kandinsky 5.0: A Family of Foundation Models for Image and Video Generation 209 upvotes, #1 of 2025-11-20
2. Reasoning via Video: The First Evaluation of Video Models' Reasoning Abilities through Maze-Solving Tasks 72 upvotes, #2 of 2025-11-20
3. What Does It Take to Be a Good AI Research Agent? Studying the Role of Ideation Diversity 54 upvotes, #3 of 2025-11-20
4. VisPlay: Self-Evolving Vision-Language Models from Images 41 upvotes, #4 of 2025-11-20
5. Instruction-Guided Lesion Segmentation for Chest X-rays with Automatically Generated Large-Scale Dataset 25 upvotes, #5 of 2025-11-20
6. ARC-Chapter: Structuring Hour-Long Videos into Navigable Chapters and Hierarchical Summaries 16 upvotes, #6 of 2025-11-20
7. MHR: Momentum Human Rig 13 upvotes, #7 of 2025-11-20
8. Mixture of States: Routing Token-Level Dynamics for Multimodal Generation 6 upvotes, #8 of 2025-11-20
9. FreeAskWorld: An Interactive and Closed-Loop Simulator for Human-Centric Embodied AI 6 upvotes, #8 of 2025-11-20
10. RoMa v2: Harder Better Faster Denser Feature Matching 6 upvotes, #8 of 2025-11-20
11. Aligning Generative Music AI with Human Preferences: Methods and Challenges 2 upvotes, #11 of 2025-11-20
12. Medal S: Spatio-Textual Prompt Model for Medical Segmentation 1 upvotes, #12 of 2025-11-20

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.