Daily Papers of 2025-08-18

1. DINOv3 194 upvotes, #1 of 2025-08-18
2. SSRL: Self-Search Reinforcement Learning 88 upvotes, #2 of 2025-08-18
3. Thyme: Think Beyond Images 76 upvotes, #3 of 2025-08-18
4. BeyondWeb: Lessons from Scaling Synthetic Data for Trillion-scale Pretraining 54 upvotes, #4 of 2025-08-18
5. XQuant: Breaking the Memory Wall for LLM Inference with KV Cache Rematerialization 39 upvotes, #5 of 2025-08-18
6. PaperRegister: Boosting Flexible-grained Paper Search via Hierarchical Register Indexing 22 upvotes, #6 of 2025-08-18
7. TexVerse: A Universe of 3D Objects with High-Resolution Textures 14 upvotes, #7 of 2025-08-18
8. FantasyTalking2: Timestep-Layer Adaptive Preference Optimization for Audio-Driven Portrait Animation 10 upvotes, #8 of 2025-08-18
9. StyleMM: Stylized 3D Morphable Face Model via Text-Driven Aligned Image Translation 9 upvotes, #9 of 2025-08-18
10. X-Node: Self-Explanation is All We Need 7 upvotes, #10 of 2025-08-18
11. Controlling Multimodal LLMs via Reward-guided Decoding 6 upvotes, #11 of 2025-08-18
12. MAESTRO: Masked AutoEncoders for Multimodal, Multitemporal, and Multispectral Earth Observation Data 4 upvotes, #12 of 2025-08-18
13. SPARSE Data, Rich Results: Few-Shot Semi-Supervised Learning via Class-Conditioned Image Translation 2 upvotes, #13 of 2025-08-18

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.