| BarunAction-35M |
| Copyright 2026 Harrrshall |
|
|
| The BarunLM source code and released model weights are licensed under the |
| Apache License 2.0. Training datasets are not redistributed with this release |
| and retain their own licenses and terms: |
|
|
| - HuggingFaceTB/smollm-corpus (FineWeb-Edu and Cosmopedia v2), revision |
| 3ba9d605774198c5868892d7a8deda78031a781f: ODC-By 1.0. FineWeb-derived |
| material is also subject to the applicable Common Crawl terms. |
| - HuggingFaceTB/finemath (FineMath-4+), revision |
| e92b25a616738fe95dc186b64dfb19f9c8525594: ODC-By 1.0 and Common Crawl |
| terms of use. |
| - mlfoundations/dclm-baseline-1.0-parquet, revision |
| 817d6752765f6a41261085171dd546b104f60626: CC BY 4.0. The dataset card |
| describes DCLM-Baseline as a research baseline. |
| - code-search-net/code_search_net (Python), revision |
| bd0cf261e357a3eb5c8fba490d23ec1a1cd59555: source repositories retain |
| their individual licenses. |
| - codeparrot/codeparrot-clean, revision |
| 35a59fb025bc0a102f7d96eac09d145b896d487b: source files retain the |
| repository-level licenses recorded by the dataset. |
|
|
| BarunAction-35M candidate-v2 was post-trained on the following additional |
| dataset, which remains under its upstream attribution terms: |
|
|
| - google/mobile-actions, revision |
| e920309bc2acbc2e99a5e3201cf37df2b9fd9151: CC BY 4.0. Candidate-v2 was |
| supervised-fine-tuned on 7,937 grouped internal-training rows; its reported |
| 756-row development population was derived only from source training rows. |
| Users are responsible for reviewing the upstream dataset cards and terms for |
| their intended use. This notice is informational and does not replace those |
| terms. |
|
|
