Datasets:

od2961
/

ModeBench

Tasks:

Text Generation

Modalities:

Formats:

Languages:

Size:

Tags:

reinforcement-learning

verifiable-rewards

Libraries:

Dataset card Data Studio Files Files and versions

ModeBench / code /oat_drgrpo /__init__.py

od2961's picture

Release frozen ModeBench Levels 1, 2 and 3 with verified splits and executable evaluation

93f2b5e verified 27 days ago

Raw History Blame Contribute Delete

71 Bytes

	"""Direct MaxEnt-GRPO primitives and the landed xDr.GRPO surrogate."""