Sakha-Judge

This is an index repo — there's no model weights here. Sakha-Judge is a benchmark/data release accompanying the paper "Sakha-Judge: A Cross-Family Benchmark for LLM-as-a-Judge Reliability on a Category-0 Language" (Egorov, Humonen).

• 🤗 Dataset & code — corpus, rubric prompts, human labels, judge scores, training + eval code
• 🧑‍⚖️ Space — landing page with resource links
• ✉️ Contact: egorovmichil9@gmail.com