This paper says we should measure an AI agent’s uncertainty across its whole conversation, not just on one final answer.
The paper introduces Entropy Sentinel, a simple way to watch how accurate an AI is by reading its “uncertainty heartbeat” during generation.
This paper teaches video-making AI models to say how sure they are about each tiny part of every frame they create.