decon
Allen Institute for AIDecontamination tool that flags and removes training documents overlapping evaluation benchmarks. The decontamination stage of the Dolma 3 pipeline.
Openness
5 high confidence- license
- Apache-2.0(OSI)
- source
- public(allenai/decon)
- core features withheld
- no
Fully OSI-licensed (Apache-2.0), full source public.
- https://github.com/allenai/decon/blob/main/LICENSE recorded 2026-08-13
Apache License Version 2.0 text
- https://raw.githubusercontent.com/allenai/decon/main/LICENSE recorded 2026-09-29
LICENSE text is the Apache License, "Apache License" "Version 2.0, January 2004", with "TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION".
- https://raw.githubusercontent.com/allenai/decon/main/README.md recorded 2026-09-29
README describes building from the published repo with "git clone https://github.com/allenai/decon" and "cargo run --release -- detect --training-dir tests/fixtures/training/", plus a server mode; searched the body for enterprise, license key, paid and premium: none.
Adoption
1 medium confidenceGitHub stars are the only usage signal published for this repository, and a star is not a use. It is a single-purpose Rust tool that covers one stage, decontamination, of the Dolma 3 pipeline behind OLMo 3, so its natural audience is small.
- https://api.github.com/repos/allenai/decon recorded 2026-08-13
stargazers_count = 36 for allenai/decon
Capability
3 medium confidenceCovers a single task, decontamination, rather than a full pipeline.
- https://raw.githubusercontent.com/allenai/decon/main/README.md recorded 2026-09-29
README states "Decon identifies documents contaminated with eval instances." and "Decon can produce contamination reports and cleaned datasets." using "token based sampling and counting methods".
Verified 2026-09-27