Tag: Benchmarking

BottleCap shrinks reasoning traces with a small accuracy trade

BottleCap AI has released a Qwen fine-tune that thinks 37.2 percent less for under a point of accuracy.