d3-lite body x-l-soup
A Zils soup body for the d3-lite pool: the full language and vision weights of a fine-tune
in the layout of Qwen/Qwen3.5-0.8B. Trained by marsplan0624 on Zils Outcomes style public decision
questions labelled by real outcomes (Wikipedia reverts and reviews, arXiv categories, Stack
Exchange outcomes, Gerrit merges) with a rank-16 language LoRA merged into the weights.
The weights are usable only through the pool's pinned chassis runtime; this repo carries no answer layer or runtime code.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support