project-themis/Themis-CodePreference
Themis-CodePreference Overview Themis-CodePreference is the largest open-source collection of code preferences to date, containing more than 350k preference pairs. It is the primary training dataset for the preference modeling (PM) stage of the Themis-RM suite of multilingual code reward models. The dataset covers five dimensions of code quality — Functional Correctness, Runtime Efficiency, Memory Efficiency, Security Hardness, and Readability &… See the full description on the dataset page: https://huggingface.co/datasets/project-themis/Themis-CodePreference.
| Author / organization | project-themis |
|---|---|
| License (as declared) | apache-2.0 |
| Downloads | 158 (as reported by Hugging Face) |
| Likes | 1 |
| Created | 30 April 2026 |
| Last updated at source | 4 May 2026 |
Popularity, last 2 days
158 now. Counts as reported by the source, snapshotted daily by Silytics.
Recent changes
- 17 Oct 2026 Description: Themis-CodePreference Overview Themis-CodePreference is the largest open-source collection of code preferences to date, containing more than 350k preference pairs. It is the primary training dataset for the preference modeling (PM) stage of → Themis-CodePreference Overview Themis-CodePreference is the largest open-source collection of code preferences to date, containing more than 350k preference pairs. It is the primary training dataset for the preference modeling (PM) stage of
Changes detected by Silytics between refreshes of the source.
More from project-themis
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.