Daily Papers of 2025-03-07

  1. START: Self-taught Reasoner with Tools 87 upvotes, #1 of 2025-03-07
  2. Token-Efficient Long Video Understanding for Multimodal LLMs 79 upvotes, #2 of 2025-03-07
  3. LLMVoX: Autoregressive Streaming Text-to-Speech Model for Any LLM 60 upvotes, #3 of 2025-03-07
  4. EgoLife: Towards Egocentric Life Assistant 35 upvotes, #4 of 2025-03-07
  5. LINGOLY-TOO: Disentangling Memorisation from Reasoning with Linguistic Templatisation and Orthographic Obfuscation 23 upvotes, #5 of 2025-03-07
  6. LLM as a Broken Telephone: Iterative Generation Distorts Information 22 upvotes, #6 of 2025-03-07
  7. Audio Flamingo 2: An Audio-Language Model with Long-Audio Understanding and Expert Reasoning Abilities 22 upvotes, #6 of 2025-03-07
  8. IFIR: A Comprehensive Benchmark for Evaluating Instruction-Following in Expert-Domain Information Retrieval 20 upvotes, #8 of 2025-03-07
  9. L^2M: Mutual Information Scaling Law for Long-Context Language Modeling 19 upvotes, #9 of 2025-03-07
  10. HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization 17 upvotes, #10 of 2025-03-07
  11. FuseChat-3.0: Preference Optimization Meets Heterogeneous Model Fusion 13 upvotes, #11 of 2025-03-07
  12. How to Steer LLM Latents for Hallucination Detection? 10 upvotes, #12 of 2025-03-07
  13. PokéChamp: an Expert-level Minimax Language Agent 9 upvotes, #13 of 2025-03-07
  14. Union of Experts: Adapting Hierarchical Routing to Equivalently Decomposed Transformer 8 upvotes, #14 of 2025-03-07
  15. Identifying Sensitive Weights via Post-quantization Integral 7 upvotes, #15 of 2025-03-07
  16. The Best of Both Worlds: Integrating Language Models and Diffusion Models for Video Generation 7 upvotes, #15 of 2025-03-07
  17. Dedicated Feedback and Edit Models Empower Inference-Time Scaling for Open-Ended General-Domain Tasks 6 upvotes, #17 of 2025-03-07
  18. Combining Flow Matching and Transformers for Efficient Solution of Bayesian Inverse Problems 5 upvotes, #18 of 2025-03-07
  19. Understanding and Predicting Derailment in Toxic Conversations on GitHub 4 upvotes, #19 of 2025-03-07
  20. Lost in Literalism: How Supervised Training Shapes Translationese in LLMs 4 upvotes, #19 of 2025-03-07
  21. On the Acquisition of Shared Grammatical Representations in Bilingual Language Models 3 upvotes, #21 of 2025-03-07

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.