Researchers · Interviewed at NeurIPS 2025
Apple

NeurIPS 2025 · December 17, 2025 · 17 min
Do today’s reasoning language models actually reason — or do they just appear to?
In this #NeurIPS 2025 highlight, Parshin Shojaee @Apple breaks down the ideas behind the widely discussed paper "The Illusion of Thinking", offering a deep dive into how modern reasoning models behave as problem complexity increases.