PVBench: Quantifying Physical Plausibility in Video Generation with Object-Level Physical Measurements
Recent video generation models produce visually compelling outputs, yet their physical plausibility remains poorly measured.
ProofPaper ↗
Key points
- Physical plausibility asks whether generated objects obey geometry, kinematics, dynamics, and conservation laws.
- Existing benchmarks rely on coarse visual-language model judgments or target embodied world models rather than video generation itself.
- PVBench comprises 600 carefully crafted prompts, of which 30% involve physical embodied agents such as robot arms, humanoids, quadrupeds, and multi-agent systems.
- By releasing the benchmark, prompts, and calibration protocol, we aim to foster physically grounded video generation.
Sources (2)
- [1]PVBench: Quantifying Physical Plausibility in Video Generation with Object-Level Physical MeasurementsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 04:44 PM
Recent video generation models produce visually compelling outputs, yet their physical plausibility remains poorly measured.
Physical plausibility asks whether generated objects obey geometry, kinematics, dynamics, and conservation laws.
- [2]PVBench: Quantifying Physical Plausibility in Video Generation with Object-Level Physical MeasurementsHugging Face Daily Papers · Oct 19, 12:00 AM · same content
Extractive summary: sentences quoted from the sources.
Before this
- Oct 18, 2026Dialogue Generation for Conversational Recommendation Using User Search Interactions and Knowledge Graphs
- Oct 17, 2026IF-MAP: Influence-Function-Inspired Agent Prioritization in Multi-Agent Systems
- Oct 17, 2026EdgeKernelBench: Unified LLM-Driven Kernel Generation for Edge Backends
- Oct 17, 2026ECHO: Efficient Causal Self-Recursion Beyond Auxiliary Multi-Token Prediction
- Oct 16, 2026Momentum: Persistent Dynamics for Video World Models
- Oct 16, 2026Persistent Planar Memory for Video World Models