Smarter model merging cuts reasoning length

A training-free method merges powerful reasoning models with concise ones, keeping accuracy while cutting output length by up to 93%.

Also today

Today's top five AI stories, one line each, at 7am Eastern. Free. One-click unsubscribe.