Microsoft releases Orca 2, a pair of small language models that outperform larger counterparts

When tested on 15 diverse benchmarks (in zero-shot settings) covering aspects like language understanding and common-sense reasoning, the Orca 2 models matched or outperformed five to ten times bigger models.

ai ai training benchmarks bigger bigger models business conversational ai diverse five generative-ai imitation learning language language models language understanding large language models llama 2 microsoft microsoft research ml and deep learning model-training nlp orca orca 2 reasoning releases science sense small small language models synthetic data understanding