# mlfoundations/dclm

DataComp for Language Models

Repository: https://github.com/mlfoundations/dclm
Canonical: https://ross.abutalabs.com/products/dclm
Language: HTML
License: MIT
License Family: permissive
Last push: 2025-09-09T04:40:36+00:00

## Health v2 (maintenance only)
Score: 42/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 41, release rhythm 35, longevity 59
- inputs: {"age_days": 828, "days_push": 358, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1467, forks 132 (observed 2026-08-28T04:04:48.741391+00:00)

## What it is
DataComp-LM (DCLM) is a benchmark and dataset suite for curating training data for language models, including a leaderboard and evaluation tooling for comparing data filtering pipelines. It provides large-scale web-crawl datasets, filtering baselines, and standardized CORE/EXTENDED evaluation scores for downstream LM performance.

## Use cases
- curate pretraining data for language models
- compare data filtering pipelines for LLM training
- evaluate dataset quality with standardized LM benchmarks
- reproduce DCLM CORE and EXTENDED scores
- build a leaderboard submission for training data selection
- filter Common Crawl data for model training

## When to choose
- you are researching or building data curation pipelines for LLM pretraining
- you need a standardized benchmark to compare dataset selection methods
- you want access to curated large-scale web-crawl training datasets

## When to avoid
- you need a general-purpose data labeling or annotation tool
- you are fine-tuning rather than pretraining and just need a small instruction dataset
- you need inference or serving infrastructure rather than training data tooling

## Facets
- artifact type: dataset
- maturity: active
- function: machine-learning, llm-training, data-science, etl, benchmarking
- domain: large-language-models, machine-learning, artificial-intelligence
- platform: python
- tags: dataset-curation, llm-pretraining, benchmark, leaderboard, data-filtering, web-crawl, data-engineering, linux, macos

## Member repositories
- mlfoundations/dclm (main) score 42

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:48.741391+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 2026-08-30T04:35:00.795259+00:00, confidence not recorded.
  - readme: https://github.com/mlfoundations/dclm (fetched 2026-08-28T04:04:48.741391+00:00, sha 000406312f8a)
- Data as of 2026-08-30T08:39:29.467469+00:00.