二进制token让图像生成快30倍，RL训练也在学会反思

重点关注

• Experiential Reinforcement Learning score 9
  入选 HF Daily Papers；HF 热度: 43 upvotes (+4)；关键词(6): efficiency, deployment, inference, agentic, reasoning
• REDSearcher: A Scalable and Cost-Efficient Framework for Long-Horizon Search Agents score 8
  入选 HF Daily Papers；HF 热度: 17 upvotes (+3)；关键词(8): efficient, agents, tool use, function calling, planning
• STATe-of-Thoughts: Structured Action Templates for Tree-of-Thoughts score 8
  入选 HF Daily Papers；HF 热度: 18 upvotes (+3)；关键词(3): inference, reasoning, search
• BitDance: Scaling Autoregressive Generative Models with Binary Tokens score 8
  入选 HF Daily Papers；HF 热度: 19 upvotes (+3)；关键词(5): scaling, inference, diffusion, multimodal, text-to-image
• UniWeTok: An Unified Binary Tokenizer with Codebook Size $\mathit{2^{128}}$ for Unified Multimodal Large Language Model score 8
  入选 HF Daily Papers；HF 热度: 10 upvotes (+3)；关键词(3): distillation, attention, multimodal
• LaViDa-R1: Advancing Reasoning for Unified Multimodal Diffusion Language Models score 6
  入选 HF Daily Papers；HF 热度: 3 upvotes (+1)；关键词(6): finetuning, post-training, diffusion, reasoning, multimodal
• LM-Lexicon: Improving Definition Modeling via Harmonizing Semantic Experts score 6
  入选 HF Daily Papers；HF 热度: 2 upvotes (+1)；关键词(2): scaling, efficient
• Fusing Pixels and Genes: Spatially-Aware Learning in Computational Pathology score 5
  关键词(2): alignment, multimodal；顶会接收: ICLR
• QuRL: Efficient Reinforcement Learning with Quantized Rollout score 5
  关键词(5): scaling, efficient, efficiency, quantization, reasoning；顶会接收: ICLR

也值得关注

• Predicting New Concept-Object Associations in Astronomy by Mining the Literature score 4
  机构: Harvard；关键词(1): inference
• HyMem: Hybrid Memory Architecture with Dynamic Retrieval Scheduling score 2
  关键词(7): efficient, efficiency, lightweight, compression, agents
• Why Code, Why Now: Learnability, Computability, and the Real Limits of Machine Learning score 2
  关键词(2): scaling, code generation
• Statistical Early Stopping for Reasoning Models score 2
  关键词(2): efficiency, reasoning
• A Generalizable Physics-guided Causal Model for Trajectory Prediction in Autonomous Driving score 2
  关键词(2): attention, agents
• A Multi-Agent Framework for Code-Guided, Modular, and Verifiable Automated Machine Learning score 2
  关键词(4): agent, agents, code generation, planning
• An Adaptive Model Selection Framework for Demand Forecasting under Horizon-Induced Degradation to Support Business Strategy and Operations score 2
  关键词(2): planning, evaluation
• A Theoretical Framework for LLM Fine-tuning Using Early Stopping for Non-random Initialization score 2
  关键词(2): fine-tuning, attention
• Eureka-Audio: Triggering Audio Intelligence in Compact Language Models score 2
  关键词(8): efficient, lightweight, MoE, reasoning, speech
• Chemical Language Models for Natural Products: A State-Space Model Approach score 2
  关键词(3): pre-training, transformer, mamba
• Steady-State Behavior of Constant-Stepsize Stochastic Approximation: Gaussian Approximation and Tail Bounds score 2
  关键词(2): scaling, efficiency
• MarsRetrieval: Benchmarking Vision-Language Models for Planetary-Scale Geospatial Retrieval on Mars score 2
  关键词(5): fine-tuning, multimodal, vision-language, benchmark, evaluation
• HLE-Verified: A Systematic Verification and Structured Revision of Humanity's Last Exam score 2
  关键词(2): benchmark, evaluation
• Neuromem: A Granular Decomposition of the Streaming Lifecycle in External Memory for LLMs score 2
  关键词(4): compression, serving, latency, cost
• DAIAN: Deep Adaptive Intent-Aware Network for CTR Prediction in Trigger-Induced Recommendation score 2
  关键词(2): real-time, recommendation