Is this your dataset? Claim it with the Hugging Face account that owns it.

HuggingFriends/
mllm-as-embodied-world-judge

Name: mllm-as-embodied-world-judge
Creator: HuggingFriends

Views

No views yet

MLLM-as-Embodied-World-Judge

Data for judging physical adherence and instruction alignment of generated embodied-manipulation videos.

	Start here

path what it is

final/ the current release — train.jsonl (11,520), test.jsonl (802), and its README

data/ source and generated videos, referenced by video_url in the splits

	Benchmark tooling

path what it is

bench/LEADERBOARD.md judge results table

bench/TESTSET.md benchmark… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFriends/mllm-as-embodied-world-judge.