Token Probabilities as Zero-Shot Rewards Hit 0.95 Correlation

Featured

• TOPReward: Token Probabilities as Hidden Zero-Shot Rewards for Robotics score 12
  机构: University of Washington; 入选 HF Daily Papers; HF 热度: 22 upvotes (+4); 关键词(5): efficiency, pretraining, vision-language, robotics, open-source
• Learning Cross-View Object Correspondence via Cycle-Consistent Mask Prediction score 10
  入选 HF Daily Papers; HF 热度: 14 upvotes (+3); 关键词(1): inference; 顶会接收: CVPR
• Anatomy of Agentic Memory: Taxonomy and Empirical Analysis of Evaluation and System Limitations score 7
  入选 HF Daily Papers; HF 热度: 5 upvotes (+2); 关键词(7): latency, throughput, agents, agentic, reasoning
• K-Search: LLM Kernel Generation via Co-Evolving Intrinsic World Model score 6
  入选 HF Daily Papers; HF 热度: 4 upvotes (+1); 关键词(4): efficient, MoE, planning, search
• SeaCache: Spectral-Evolution-Aware Cache for Accelerating Diffusion Models score 5
  关键词(3): inference, latency, diffusion; 顶会接收: CVPR
• A Benchmark and Knowledge-Grounded Framework for Advanced Multimodal Personalization Study score 5
  机构: Google; 关键词(3): reasoning, multimodal, benchmark
• Towards Calibrating Prompt Tuning of Vision-Language Models score 5
  关键词(2): efficient, vision-language; 顶会接收: CVPR
• IDLM: Inverse-distilled Diffusion Language Models score 5
  机构: Huawei; 关键词(3): distillation, inference, diffusion
• ChordEdit: One-Step Low-Energy Transport for Image Editing score 5
  关键词(5): fast, lightweight, inference, real-time, text-to-image; 顶会接收: CVPR
• Universal 3D Shape Matching via Coarse-to-Fine Language Guidance score 5
  关键词(2): attention, multimodal; 顶会接收: CVPR

Also Worth Noting

• Ani3DHuman: Photorealistic 3D Human Animation with Self-guided Stochastic Sampling score 4
  入选 HF Daily Papers; 关键词(1): diffusion
• MRI Contrast Enhancement Kinetics World Model score 4
  关键词(1): alignment; 顶会接收: CVPR
• Stable Deep Reinforcement Learning via Isotropic Gaussian Representations score 4
  机构: Mila; 关键词(1): agents
• Why ReLU? A Bit-Model Dichotomy for Deep Network Training score 3
  机构: MIT
• Online Realizable Regression and Applications for ReLU Networks score 3
  机构: MIT
• InfEngine: A Self-Verifying and Self-Optimizing Intelligent Engine for Infrared Radiation Computing score 2
  关键词(2): agents, coding
• Quantifying Automation Risk in High-Automation AI Systems: A Bayesian Framework for Failure Propagation and Optimal Oversight score 2
  关键词(3): efficient, deployment, agentic
• IDSelect: A RL-Based Cost-Aware Selection Agent for Video-based Multi-Modal Person Recognition score 2
  关键词(5): efficiency, lightweight, inference, agent, cost
• Benchmark Test-Time Scaling of General LLM Agents score 6
  入选 HF Daily Papers; HF 热度: 2 upvotes (+1); 关键词(8): scaling, agent, agents, coding, reasoning
• MagicAgent: Towards Generalized Agent Planning score 2
  关键词(6): lightweight, fine-tuning, agent, agents, planning
• GUIDE-US: Grade-Informed Unpaired Distillation of Encoder Knowledge from Histopathology to Micro-UltraSound score 2
  关键词(2): distillation, inference
• Capable but Unreliable: Canonical Path Deviation as a Causal Mechanism of Agent Failure in Long-Horizon Tasks score 2
  关键词(4): scaling, agent, agents, benchmark
• Learning to Detect Language Model Training Data via Active Reconstruction score 5
  入选 HF Daily Papers; 关键词(5): distillation, inference, finetuning, post-training, pre-training
• FedAvg-Based CTMC Hazard Model for Federated Bridge Deterioration Assessment score 2
  关键词(4): planning, benchmark, evaluation, synthetic data
• Routing-Aware Explanations for Mixture of Experts Graph Models in Malware Detection score 2
  关键词(4): edge, MoE, mixture of experts, reasoning