Episode Details

Today's papers: - Beyond Inference-Time Search: Reinforcement Learning Synthesizes Reusable Solvers: https://arxiv.org/abs/2605.18374v1 - The Hidden Cost of Contextual Sycophancy: an AI Literacy Intervention in Human-AI Collaboration: https://arxiv.org/abs/2605.18372v1 - CommitDistill: A Lightweight Knowledge-Centric Memory Layer for Software Repositories: https://arxiv.org/abs/2605.18284v1 - Training data attribution in diffusion models via mirrored unlearning and noise-consistent skew: https://arxiv.org/abs/2605.17938v1 - Predictable Confabulations: Factual Recall by LLMs Scales with Model Size and Topic Frequency: https://arxiv.org/abs/2605.18732v1 This podcast is from Colin Davis (colin-davis.com) using Claude & Elevenlabs.