Mtrini-Decision-1.0 research preview

An experimental decision-model checkpoint (training step 100): a Qwen3-8B LoRA adapter with a learned decision head for structured choices and option ranking. No validated benchmark result is claimed.