clm-projects/jev-boost-kn-reuters-sata-qwen3.6-35b-a3b-fp8-embeddings

Jev boost: kn-reuters-sata: Qwen3.6-35B-A3B-FP8 state/action embeddings ucirvine/reuters21578 (revision 321cce2a6934016cebbb9708894bf86cb46edb90), split train, converted into decisions shaped like the Jev Decision Index's Knowledge & Reasoning requests (SATA-Bench) and embedded with Qwen/Qwen3.6-35B-A3B-FP8 (vLLM pooling, last token, L2-normalized, 2048-d) for post-training CLM heads (clm_train: clm_train.data.jev_boost, docs/jev-boost.md). ModApte train news x topics (gold +… See the full description on the dataset page: https://huggingface.co/datasets/clm-projects/jev-boost-kn-reuters-sata-q

<iframe src="https://silytics.si/embed/item/1863be0d-5375-4954-b563-78173f98ea55" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 24 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.