MUTUAL REASONING MAKES SMALL MODELS SMARTER

Summary from the TLDR AI newsletter

Tree search is an extremely active area of research for inference time computation for language models. This paper from Microsoft shows a particularly compelling case that enables small models to improve dramatically on mathematics problems.