1 articles with this tag
Alejandro Vidal argues for adopting psychometric principles like IRT to move beyond outdated LLM benchmarking, enabling more nuanced and reliable model evaluation.
