A human-scored benchmark of 84 physics-law prompts finds all ten tested text-to-video models average below 0.42, indicating weak physical consistency.
Genai-bench: A holistic benchmark for com- positional text-to-visual generation
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
T2VPhysBench: A First-Principles Benchmark for Physical Consistency in Text-to-Video Generation
A human-scored benchmark of 84 physics-law prompts finds all ten tested text-to-video models average below 0.42, indicating weak physical consistency.