Hugging Face's logo

Datasets:
alphaXiv
/
scope-rl-reproduction-data

Dataset card Data Studio Files
xet
 Community
scope-rl-reproduction-data / README.md
alphaXiv's picture
alphaXiv
Upload README.md with huggingface_hub
a7d9d3c verified
preview code
|
Raw
History Blame Contribute Delete
425 Bytes

SCOPE-RL reproduction data

Prepared from the SCOPE-RL paper repo (github.com/tokencraft-lab/SCOPE-RL, arXiv 2607.11506) by scope_rl/prep_data.py in github.com/alphaXiv/paper-implementations.

  • dapo_math_2400.jsonl - DAPO-Math 2400 training subset: chat prompt + rule-verifiable ground truth (sub-question decompositions stripped)
  • benchmark.jsonl - paper eval benchmark: AIME24/25 (x8 rows), MATH500, GPQA-diamond