zils-lite-f1

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.

Start: the d3-lite soup of round d3-lite-2026-11-06, the uniform weight average of 1lock/zils-lite-astrid-g1 at revision ac979f3d and bmars8218/zils-lite-body-2 at revision 60b596b3 (both Apache-2.0). Four fine-tunes of that start were trained independently and their weights averaged uniformly: two full fine-tunes of every language layer (lr 4e-6, cosine schedule, about 200k rows each, cross-entropy + Brier loss, with a KL anchor to the start on 40k broad rows) and two one-epoch LoRA runs (rank 32, alpha 64, lr 1e-4, merged) on 4.1k exam-format items.

Training data: 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 training splits only, plus the public Zils relay teacher answers. No validator exam items and no public seed items were used. This repository holds weights only; check the source datasets' terms before reuse.