Tiny MLP circuit makes LLMs overthink before stopping
Mechanistic interpretability work on Qwen3‑1.7B shows that a handful of late‑layer MLPs act as a "termination circuit" that fires the </think> token once the model's written answer matches its internal computation. The answer is usually settled after only ~30% of the chain‑of‑thought, meaning most of the generated reasoning is overkill and hard to steer.