Attributions
Attributions
The public datasets behind the benchmark board — what each one is, where it comes from, and the licence it carries. Model artefacts (the encoder heads and the per-domain specialists) are first-party: built and trained by us, never a redistribution of someone else's weights.
| Board suite | Source | Licence | Notes |
|---|---|---|---|
| Typed decisions | LocalLLaMA/typed-decisions | Apache-2.0 | First-party synthetic benchmark. |
| MASSIVE intent | mteb/amazon_massive_intent — a mirror of Amazon's MASSIVE | CC BY 4.0 | © Amazon.com Inc. or its affiliates. Cite FitzGerald et al. 2022. |
| Banking77 | mteb/banking77 — a mirror of PolyAI's BANKING77 | CC BY 4.0 | © PolyAI. Cite Casanueva et al. 2020. |
| Prompt injections | deepset/prompt-injections | Apache-2.0 | — |
| Thai Wisesight | pythainlp/wisesight_sentiment | CC0 | — |
| Thai Sib200 | Davlan/sib200 | CC BY-SA 4.0 | Share-alike: derivatives that redistribute it stay share-alike. |
Mirrors and upstream
Where a suite names a mirror, the licence is the upstream data owner's, not the mirror repository's — the attributions above follow the upstream licence.
The two Thai suites
The two Thai suites are recorded-only benchmarks, not served lanes.
First-party artefacts
Model artefacts — encoder heads, per-domain specialists, and our own synthetic corpora — are first-party.
Contact
Operated by gist.rs. Licensing questions: hello@gist.rs.