This podcast episode features Nathan Lambert and Sebastian Raschka discussing Anthropic's distillation techniques and how models can cheat on benchmarks. The conversation also touches upon the SWE-Bench benchmark, indicating it may be defunct. The episode was part of SAIL Live #6 and is available to paid subscribers of The Latent Space podcast. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON This is a podcast episode featuring discussion by named credible voices on AI topics, fitting the commentary bucket.
