LockedInTheSkage/

RoundixReason-tagged-reasoning

Multi-domain reasoning dataset for GRPO training.

	Domains




	

Domain Source

math MetaMathQA-40K

science ScienceQA, SciQ

code LeetCodeDataset

creative_writing WritingPrompts

{ "question": str, # Input prompt "answer": str, # Expected response "domain": str, # One of: math, science, code, creative_writing }

Code examples include additional fields: test_code, entry_point, helpers.