Meetup: Evaluating LLMs: Needle in a Haystack

LLM evaluation is a discipline where confusion reigns and foundation model builders are effectively grading their own homework.

​Building on the viral threads on X/Twitter, Greg Kamradt, Robert Nishihara, and Jason Lopatecki discuss highlights from Arize AI's ongoing research on how major foundation models – from OpenAI’s GPT-4 to Mistral and Anthropic’s Claude – are stacking up against each other at important tasks and emerging LLM use cases, covering and explaining the importance of results of Needle in a Haystack tests and other evals results on hallucination detection on private data, question-and-answer, code functionality, and more.

​Curious which foundation models your company should be using for a specific use case – and which to avoid? You won’t want to miss this meetup!

[antonidabrowski]

Good content, thanks for your research