d3-mini decision body
A decision body for the Zils d3-mini soup pool (Bittensor subnet 127): Qwen3.5-4B layout (language and vision tensors, bf16).
Starting weights: the uniform average of torchcast-ai/torchcast-decision-4b and vllm-sr/d3-mini (both Apache-2.0). Four LoRA runs (rank 16, one epoch) on different data slices were merged and averaged with equal weights.
Training data: distillation targets from the public zils-ai/zils-relay-answers dataset (Apache-2.0), synthetic decision questions generated by our own code (counting, attribute coverage, tool-call policy, word forms), and public train splits (CMUdict train, 20 Newsgroups train, PIQA train, STS-B train), plus regularisation toward the starting weights. No validator exam data was used.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for BigDok/d3-mini-body-1
Base model
Qwen/Qwen3.5-4B-Base Finetuned
Qwen/Qwen3.5-4B Finetuned
torchcast-ai/torchcast-decision-4b