PhysLoop: Harnessing LLMs' Physics Awareness for Self-Improving Video Generation Models
cs.CV, cs.AI
Submitted: 2026-11-12
Updated: 2026-11-15
Code: https://github.com/DAGroup-PKU/PhysLoop
Project page: https://chenyx24.github.io/PhysLoop-site
Terminology
Sources
- ComPhy: Compositional Physical Reasoning of Objects and Events from Videos
- NEWTON: Agentic Planning for Physically Grounded Video Generation
- Improving Dynamic Object Interactions in Text-to-Video Generation with AI Feedback
- VBench: Comprehensive Benchmark Suite for Video Generative Models
- Self-Refining Video Sampling
- SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning
- HunyuanVideo: A Systematic Framework For Large Video Generative Models
- Self-Correcting Text-to-Video Generation with Misalignment Detection and Localized Refinement
- PhysAgent: Reflective Agentic Physics Control for Physically Plausible Video Generation
- Scene-Level Heterogeneous Physics Simulation with 3D Gaussian Splats
- Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation
- Movie Gen: A Cast of Media Foundation Models
- Physics-IQ Verified
- SimWorld: An Open-ended Realistic Simulator for Autonomous Agents in Physical and Social Worlds
- VideoAgent: Self-Improving Video Generation
- CLEVRER: CoLlision Events for Video REpresentation and Reasoning
- Vision-as-Inverse-Graphics Agent via Interleaved Multimodal Reasoning
- GS-Agent: Creating 4D Physical Worlds With Generative Simulation
- Physion-Eval: Evaluating Physical Realism in Generated Video via Human Reasoning
- VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulness
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs