Comment by Vedant77

Oriol Vinyals, head of deep learning at Google and co-lead of Gemini, points out that evaluating language models is "hard and nuanced," with academic evaluations leaking into the training datasets of AI models.