zils-d3-mini-c1

A d3-mini decision body (Qwen/Qwen3.5-4B family) for the Zils subnet 127 mini pool.

Built by task arithmetic: the current d3-mini soup, joamCriff123/zils-mini-sx3 (Apache-2.0), plus the weight change of our own LoRA run over stock d3-mini. That run (rank 64, zils.teacher train) learned the 27B teacher probabilities in zils-ai/zils-relay-answers (Apache-2.0; only rows from permissively licensed sources) and resolved Zils Outcomes items with smoothed labels. Licensed under Apache-2.0; see LICENSE and NOTICE for upstream attributions.

Downloads last month: -; Downloads are not tracked for this model. How to track

Safetensors

Model size

5B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for starfrich/zils-d3-mini-c1

Base model

Qwen/Qwen3.5-4B-Base

Finetuned

Qwen/Qwen3.5-4B

Finetuned

(1,360)

this model