zils-lite-qb3
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 joamCriff123/zils-lite-l1117a, the sole body of the 18 November 2026 d3-lite soup (Apache-2.0); credit for the start belongs to its author and to the authors of the bodies it was built from. Procedure: three LoRA (rank 32, merged) fine-tunes of that body with different seeds, one epoch each at learning rate 5e-5, were averaged uniformly; no interpolation. Training targets were soft: for most written items one quarter the item's own answer and three quarters the answer probabilities of joamCriff123/zils-mini-m1115 (the sole body of the 18 November 2026 d3-mini soup; teacher distillation from the larger pool); for the about 4,100 written arithmetic-score and sentence-similarity score items that body's probabilities alone, because their written answers proved unreliable; for the 15,000 unlabeled rows that body's probabilities only; for the 30,000 labelled rows one half the row's own dataset label and one half the starting body's own answer probabilities.
Training data (about 76,100 rows, one shuffled pass per seed): (1) about 31,100 exam-format decision items of the known task families written with the public Zils authoring recipe (scripts/author_exam.py) by Claude Opus and by a local open model, from train-side splits of the public source datasets, including about 1,100 tool-call items written from rows 90,000-93,000 of the function-call subset of glaiveai/glaive-function-calling-v2 (a dataset without splits that the exam recipe also draws from); (2) 15,000 rows of the train split of tasksource/tasksource-jev-typed-decisions used only as unlabeled distillation inputs (labels not used; sources with commercial or unspecified licence metadata, non-commercial sources excluded); (3) 30,000 further rows of the train split of tasksource/tasksource-jev-typed-decisions from 240 other source tasks whose licence metadata is unspecified, used with their labels (non-commercial sources excluded; sources named like BIG-bench tasks dropped; rows whose text equals a validation or test row dropped). 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.