I broke Meta’s Llama 3.1 405B with one question (which GPT-4o mini gets right)

The failure of the large language model may point to issues with the broad use of synthetic data.

Source: Latest news