Datasets:

rl-actors
/

Big-Math-RL-Verified-Subset

dataset_info:
  features:
    - name: problem
      dtype: string
    - name: answer
      dtype: string
    - name: source
      dtype: string
    - name: domain
      sequence: string
    - name: llama8b_solve_rate
      dtype: float64
  splits:
    - name: train
      num_bytes: 4885007.599015618
      num_examples: 15938
    - name: test
      num_bytes: 542812.6777422926
      num_examples: 1771
  download_size: 2823831
  dataset_size: 5427820.276757911
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: test
        path: data/test-*
license: apache-2.0
task_categories:
  - question-answering
language:
  - en
tags:
  - math
  - RL
size_categories:
  - 10K<n<100K

Default Math dataset used for https://github.com/RD211/actors.

The dataset is a subset of https://huggingface.co/datasets/SynthLabsAI/Big-Math-RL-Verified.

Only includes problems with solve rate between 0.05 and 0.6 and from aops_forum and olympiads.