HF RL Explorer

OctoReasoner/verl_py_sgrep_aggressive_filtered

verl py sgrep aggressive filtered: RL dataset on Hugging Face. Curriculum-filtered subset of the first 30,000 rows of OctoReasoner/verl py sgrep aggressive. CoT variant (models prompted to think first, then box; max tokens=4096). Kept rows satisfy weak <= 5/8 AND strong = 2/8, where weak/strong =…

OctoReasoner/verl_py_sgrep_aggressive_filtered on the Hugging Face Hub