RL Patches 3D Consistency Into Video Models Without Touching Architecture

Featured

• World-R1: Reinforcing 3D Constraints for Text-to-Video Generation score 13
  机构: Microsoft; 入选 HF Daily Papers; HF 热度: 110 upvotes (+4); 有代码实现; 关键词(3): GRPO, vision-language, text-to-video
• Rewarding the Scientific Process: Process-Level Reward Modeling for Agentic Data Analysis score 7
  入选 HF Daily Papers; HF 热度: 17 upvotes (+3); 关键词(3): scaling, agentic, reasoning

Also Worth Noting

• LearnPruner: Rethinking Attention-based Token Pruning in Vision Language Models score 4
  关键词(3): pruning, reasoning, vision-language; 顶会接收: ICLR
• Hindsight Preference Optimization for Financial Time Series Advisory score 4
  关键词(3): DPO, reasoning, vision-language; 顶会接收: ICLR
• TopoHR: Hierarchical Centerline Representation for Cyclic Topology Reasoning in Driving Scenes with Point-to-Instance Relations score 4
  关键词(1): reasoning; 顶会接收: CVPR
• Don't Pause! Every prediction matters in a streaming video score 4
  机构: Yale; 关键词(4): scaling, latency, real-time, post-training
• DPEPO: Diverse Parallel Exploration Policy Optimization for LLM-based Agents score 4
  关键词(2): fine-tuning, reasoning; 顶会接收: ACL
• Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs score 4
  关键词(2): DPO, vision-language; 顶会接收: ACL
• BMD-45: A Large-Scale CCTV Vehicle Detection Dataset for Urban Traffic in Developing Cities score 4
  关键词(1): deployment; 顶会接收: CVPR
• PhysNote: Self-Knowledge Notes for Evolvable Physical Reasoning in Vision-Language Model score 4
  关键词(3): agentic, reasoning, vision-language; 顶会接收: ICLR
• Zero-shot Large Language Models for Automatic Readability Assessment score 4
  关键词(1): open-source; 顶会接收: ACL
• Aligned Multi-View Scripts for Universal Chart-to-Code Generation score 4
  关键词(3): lightweight, code generation, open-source; 顶会接收: ACL
• Meta-CoT: Enhancing Granularity and Generalization in Image Editing score 13
  机构: Meta; 入选 HF Daily Papers; HF 热度: 24 upvotes (+4); 有代码实现; 关键词(1): reasoning
• The Price of Agreement: Measuring LLM Sycophancy in Agentic Financial Applications score 4
  关键词(1): agentic; 顶会接收: ICLR
• Can LLMs Act as Historians? Evaluating Historical Research Capabilities of LLMs via the Chinese Imperial Examination score 4
  关键词(1): reasoning; 顶会接收: ACL
• CT-FineBench: A Diagnostic Fidelity Benchmark for Fine-Grained Evaluation of CT Report Generation score 3
  顶会接收: ACL
• Jailbreaking Frontier Foundation Models Through Intention Deception score 3
  顶会接收: CVPR