AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Generative assessment project: Comparing language models on hallucinations and answering behavior
OpenAI unveiled the generative assessment project, a research initiative to explore strengths and weaknesses of language models. They compare how different models hallucinate answers and hedge instead of giving a direct answer. It's interesting because it's a unique comparison that hasn't been done before, especially with proprietary models.