4 ms·New technique improves the reasoning capabilities of large language models2 points by wglb 2y ago