The Termination Circuit (how reasoning models stop thinking).

·LessWrong··

Reasoning models since the dawn of o1 and R1 have a tendency to overthink. Despite a lot of work on early-exit methods and steering, open-weight and smaller reasoning models still produce long chains of thought before they answer. I worked on discovering how much of that thinking is required before the model already knows the answer and what makes the model stop thinking.SummaryReasoning models keep thinking for a long time even after finding the answer.In Qwen3-1.7B, the answer to most GSM8K pr...

Read full article →

Related Articles

GPT-5.6 Sol Ultra produces proof of the Cycle Double Cover Conjecture [pdf]
scrlk · Hacker News · 8h ago
How the terrorist group Boko Haram uses frontier AI
imustachyou · Hacker News · 8h ago
QuadRF can spot drones and see WiFi through my wall
speckx · Hacker News · 11h ago
The glass backbone: Why the Army's logistics will break in the next war
baud147258 · Hacker News · 1d ago
Snails' teeth beats spider silk as nature's strongest material (2015)
simonebrunozzi · Hacker News · 10h ago