adamjb111/nla-qwen36-27b-L42-distill-data

Qwen3.6-27B layer 28 / 42 / 52 activations with NLA explanations Inputs for distilling the WorkspaceBench NLA ceselder/qwen3.6-27b-nla-rl (av_base + av_rl_adapters/iter_000400, layer 42) into small Qwen3.5 students. Text: the teacher's own training prefixes (FineFineWeb, from ceselder/qwen3-8b-nla-L24-finefineweb-100k), 263k rows from 87.7k docs, up to 3 per doc, in a fixed doc order so in_33k / in_66k / in_131k / in_263k subsets nest. heldout: 1,020 prefixes from 340 fresh… See the full description on the dataset page: https://huggingface.co/datasets/adamjb111/nla-qwen36-27b-L42-distill-data.

<iframe src="https://silytics.si/embed/item/b471e00c-179c-41ba-b2b6-332bf6d41e10" 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 Information may be incomplete or out of date.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 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.