Daily Papers of 2025-09-05

  1. Drivel-ology: Challenging LLMs with Interpreting Nonsense with Depth 198 upvotes, #1 of 2025-09-05
  2. From Editor to Dense Geometry Estimator 86 upvotes, #2 of 2025-09-05
  3. Towards a Unified View of Large Language Model Post-Training 67 upvotes, #3 of 2025-09-05
  4. Inverse IFEval: Can LLMs Unlearn Stubborn Training Conventions to Follow Real Instructions? 54 upvotes, #4 of 2025-09-05
  5. DeepResearch Arena: The First Exam of LLMs' Research Abilities via Seminar-Grounded Tasks 53 upvotes, #5 of 2025-09-05
  6. NER Retriever: Zero-Shot Named Entity Retrieval with Type-Aware Embeddings 28 upvotes, #6 of 2025-09-05
  7. Transition Models: Rethinking the Generative Learning Objective 28 upvotes, #6 of 2025-09-05
  8. Loong: Synthesize Long Chain-of-Thoughts at Scale through Verifiers 22 upvotes, #8 of 2025-09-05
  9. Video-MTR: Reinforced Multi-Turn Reasoning for Long Video Understanding 17 upvotes, #9 of 2025-09-05
  10. Few-step Flow for 3D Generation via Marginal-Data Transport Distillation 11 upvotes, #10 of 2025-09-05
  11. Durian: Dual Reference-guided Portrait Animation with Attribute Transfer 9 upvotes, #11 of 2025-09-05
  12. Drawing2CAD: Sequence-to-Sequence Learning for CAD Generation from Vector Drawings 8 upvotes, #12 of 2025-09-05
  13. Delta Activations: A Representation for Finetuned Large Language Models 5 upvotes, #13 of 2025-09-05
  14. False Sense of Security: Why Probing-based Malicious Input Detection Fails to Generalize 2 upvotes, #14 of 2025-09-05

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.