INSAIT-Institute/ImplausiBench

🧠 ImplausiBench ImplausiBench is a benchmark for evaluating physical plausibility understanding in video–language models.It contains 150 paired real and implausible video scenarios (300 total), each with a structured multiple-choice Q/A prompt testing whether a model can detect physically impossible events. Dataset Summary Each entry contains: A question about whether the scene follows physics principles. Seven options (A–G) mixing plausible and implausible… See the full description on the dataset page: https://huggingface.co/datasets/INSAIT-Institute/ImplausiBench.

<iframe src="https://silytics.si/embed/item/8bf0ddf3-0cee-442a-8912-d43d72a3d7ab" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.