Physion-Eval: Evaluating Physical Realism in Generated Video via Human Reasoning

arXiv 2026 43.5 benchmark

TLDR

Physion-Eval benchmark uses expert human reasoning to evaluate physical realism failures in generated videos, finding most state-of-the-art generated videos contain physical glitches.

Reasoning

The paper introduces a large-scale expert-annotated benchmark with 10,990 reasoning traces and fine-grained physical categories, providing strong evidence of current video generation limitations. Its weakness is that it focuses on human-identifiable glitches and does not propose automated metrics or mitigation strategies.

Read-first score

Read-first score 43.5, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 18.

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,dataset,evaluation,metric

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=dataset

Topical relevance 42%

25.7

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 667.

Keyword Scores

world model

world simulator

video world model

generative world model

world dynamics prediction

interactive world model

model-based reinforcement learning world model

Resources

Paper