Topic
Out Of Distribution Detection
2 episodes
-
Overview: Out of Distribution Detection
We finally slow down and explain out-of-distribution detection properly — what it is, why overconfident models are actually the core danger, and how the field tries to give models a sense of where their knowledge ends.
-
Tracing Agentic Failure from the Flow of Success
Jessica and Cathy dig into Oat, a lightweight unsupervised failure-attribution model for agentic systems that learns only from successful trajectories. They unpack why the paper matters for debugging long-horizon agents, how Neural CDEs and a gated control path turn success traces into a normal-flow model, and why the speed and no-label setup make it feel more shippable than prompt-heavy baselines.