AI Thinking to recall: How reasoning unlocks parametric knowledge in LLMs Google Research · 24 June 2026 at 18:51 · 5 min read Read in Synapse Read on research.google