zils-lite-lnj1
A body for the Zils d3-lite soup pool (Bittensor subnet 127): a fine-tune of Qwen/Qwen3.5-0.8B in the chassis layout, BF16, language weights trained, vision tower from the parent.
Soup fine-tune: the start is the uniform weight average of the two bodies of the 14 November 2026 d3-lite soup (VadymPikulyk/zils-lite-dh2 and qhoenix/zils-lite-decision-avg-abc), both Apache-2.0; credit for the start belongs to their authors and to the authors of the bodies they were built from. The body is the one rank-64 LoRA run (lr 5e-5, about one epoch) on the training data below, merged into the start.
Training data (about 109,000 rows): (1) exam-format decision items written with the public Zils authoring recipe (scripts/author_exam.py) by Claude Opus and by a local open model, from public dataset rows: train-side splits for the known task families; for the new-domain style tasks, rows of the public new-domain task pools listed in scripts/author_exam.py (BIG-bench task files and splits of other public datasets, mostly held-out validation/test splits; BIG-bench tasks used by the public seed exams were excluded); (2) 63,840 typed decisions from the train split of tasksource/tasksource-jev-typed-decisions (only rows marked license_use: commercial, mostly-ASCII text, at most 200 rows per source dataset, 413 sources), converted to the Zils training format; each row keeps its source's licence, see that dataset's card. No validator exam items, no public seed exam items and no customer data were used. This repository holds weights only; check the source datasets' terms before reuse.