Created on November 03, 2026
2026
New preprint on arXiv: AGORA, reliability-aware aggregation of LLM judges via item-response theory.