Reinforcement learning helps AI models express uncertainty honestly

Original: Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs

Source: arxiv.org ↗