Daily Papers of 2025-04-21

  1. Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model? 113 upvotes, #1 of 2025-04-21
  2. NodeRAG: Structuring Graph-based RAG with Heterogeneous Nodes 41 upvotes, #2 of 2025-04-21
  3. MIG: Automatic Data Selection for Instruction Tuning by Maximizing Information Gain in Semantic Space 36 upvotes, #3 of 2025-04-21
  4. Could Thinking Multilingually Empower LLM Reasoning? 27 upvotes, #4 of 2025-04-21
  5. AerialMegaDepth: Learning Aerial-Ground Reconstruction and View Synthesis 20 upvotes, #5 of 2025-04-21
  6. It's All Connected: A Journey Through Test-Time Memorization, Attentional Bias, Retention, and Online Optimization 17 upvotes, #6 of 2025-04-21
  7. Analyzing LLMs' Knowledge Boundary Cognition Across Languages Through the Lens of Internal Representations 17 upvotes, #6 of 2025-04-21
  8. Generative AI Act II: Test Time Scaling Drives Cognition Engineering 16 upvotes, #8 of 2025-04-21
  9. CLASH: Evaluating Language Models on Judging High-Stakes Dilemmas from Multiple Perspectives 14 upvotes, #9 of 2025-04-21
  10. HiScene: Creating Hierarchical 3D Scenes with Isometric View Generation 11 upvotes, #10 of 2025-04-21
  11. Tokenize Image Patches: Global Context Fusion for Effective Haze Removal in Large Images 10 upvotes, #11 of 2025-04-21
  12. Thought Manipulation: External Thought Can Be Efficient for Large Reasoning Models 8 upvotes, #12 of 2025-04-21
  13. Self-alignment of Large Video Language Models with Refined Regularized Preference Optimization 4 upvotes, #13 of 2025-04-21
  14. Cost-of-Pass: An Economic Framework for Evaluating Language Models 4 upvotes, #13 of 2025-04-21
  15. Filter2Noise: Interpretable Self-Supervised Single-Image Denoising for Low-Dose CT with Attention-Guided Bilateral Filtering 1 upvotes, #15 of 2025-04-21
  16. Revisiting Uncertainty Quantification Evaluation in Language Models: Spurious Interactions with Response Length Bias Results 1 upvotes, #15 of 2025-04-21

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.