Today's papers:

- Optimization before Evaluation: Evaluation with Unoptimised Prompts Can be Misleading: https://arxiv.org/abs/2604.27637v1

- Position: agentic AI orchestration should be Bayes-consistent: https://arxiv.org/abs/2605.00742v1

- From Context to Skills: Can Language Models Learn from Context Skillfully?: https://arxiv.org/abs/2604.27660v1

- CastFlow: Learning Role-Specialized Agentic Workflows for Time Series Forecasting: https://arxiv.org/abs/2604.27840v1

- CoAX: Cognitive-Oriented Attribution eXplanation User Model of Human Understanding of AI Explanations: https://arxiv.org/abs/2604.27354v1

This podcast is from Colin Davis (colin-davis.com) using Claude & Elevenlabs.