# openai/prm800k

800,000 step-level correctness labels on LLM solutions to MATH problems

Repository: https://github.com/openai/prm800k
Canonical: https://ross.abutalabs.com/products/prm800k
Language: Python
License: MIT
License Family: permissive
Archived: true
Last push: 2023-06-01T17:41:14+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 88
- inputs: {"age_days": 1238, "days_push": 1189, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 2150, forks 129 (observed 2026-08-28T04:06:19.295216+00:00)

## What it is
PRM800K is OpenAI's process supervision dataset containing 800,000 step-level correctness labels for model-generated solutions to MATH dataset problems. It accompanies the paper 'Let's Verify Step by Step' and includes raw labels plus labeler instructions.

## Use cases
- train a process reward model for math reasoning
- get step-level correctness labels for LLM solutions
- research process supervision versus outcome supervision
- fine-tune LLMs on human-labeled reasoning steps
- evaluate LLM mathematical problem solving
- study human feedback data collection for RLHF

## When to choose
- you need step-level human feedback data for training reward models
- you are researching process supervision for mathematical reasoning
- you want a benchmark dataset for LLM solution correctness

## When to avoid
- you need a ready-to-use reward model rather than raw labels
- you need labels for domains other than math problems
- you want actively maintained tooling rather than a static dataset

## Facets
- artifact type: dataset
- maturity: maintenance
- function: machine-learning, llm-training, data-science
- domain: large-language-models, machine-learning, mathematics
- platform: python, cross-platform
- tags: process-supervision, reward-models, mathematical-reasoning, step-level-labels, llm-evaluation, human-annotation, natural-language-processing

## Member repositories
- openai/prm800k (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:19.295216+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T02:50:58.029194+00:00, confidence not recorded.
  - readme: https://github.com/openai/prm800k (fetched 2026-08-28T04:06:19.295216+00:00, sha e242eea46440)
- Data as of 2026-08-30T08:39:29.467469+00:00.