Dataset assetOpen Source CommunityMachine LearningNatural Language Processing

prm800k

This dataset contains data from [openai/prm800k](https://github.com/openai/prm800k). It is divided into two phases (phase1 and phase2), each with train and test splits. Features include labeler, timestamp, question, etc.; detailed feature types are described in the README.

Source

huggingface

Created

Dec 13, 2024

Updated

Dec 14, 2024

Signals

515 views

Availability

Linked source ready

Overview

Dataset description and usage context

Dataset Overview

Dataset Information

Configuration `phase1`

Features:
- labeler: string
- timestamp: string
- generation: null
- is_quality_control_question: bool
- is_initial_screening_question: bool
- question (structured):
  - problem: string
  - ground_truth_answer: string
- label (structured):
  - steps (list):
    - completions (list):
      - text: string
      - rating: int64
      - flagged: bool
    - human_completion (structured):
      - text: string
      - rating: null
      - source: string
      - flagged: bool
      - corrected_rating: int64
    - chosen_completion: int64
  - total_time: int64
  - finish_reason: string
Splits:
- train: 5,185,121 bytes, 949 samples
- test: 532,137 bytes, 106 samples
Download Size: 1,850,110 bytes
Dataset Size: 5,717,258 bytes

Configuration `phase2`

Features:
- labeler: string
- timestamp: string
- generation: int64
- is_quality_control_question: bool
- is_initial_screening_question: bool
- question (structured):
  - problem: string
  - ground_truth_solution: string
  - ground_truth_answer: string
  - pre_generated_steps: sequence of string
  - pre_generated_answer: string
  - pre_generated_verifier_score: float64
- label (structured):
  - steps (list):
    - completions (list):
      - text: string
      - rating: int64
      - flagged: bool
    - human_completion: null
    - chosen_completion: int64
  - total_time: int64
  - finish_reason: string
Splits:
- train: 344,736,273 bytes, 97,782 samples
- test: 9,164,167 bytes, 2,762 samples
Download Size: 132,668,705 bytes
Dataset Size: 353,900,440 bytes

Configuration Files

phase1:
- train: phase1/train-*
- test: phase1/test-*
phase2:
- train: phase2/train-*
- test: phase2/test-*

Language

English (en)

Dataset Scale

10K < n < 100K

Need downstream help?

Pair the dataset with AI analysis and content workflows.

Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.

Explore AI studio