Orca 2-7b Model

About Model

Orca-2-7B is a research-focused language model developed by Microsoft, fine-tuned from Meta's LLaMA-2 7B base model. It aims to improve the reasoning abilities of smaller language models by utilizing a synthetic dataset specifically curated for this purpose. The model excels in single-turn tasks, including reasoning over user-provided data, reading comprehension, mathematical problem-solving, and text summarization. Notably, Orca-2-7B has demonstrated performance comparable to much larger models, achieving significant improvements in complex zero-shot reasoning benchmarks. However, it is not optimized for chat applications and has not undergone reinforcement learning from human feedback (RLHF) or Direct Preference Optimization (DPO). Researchers are encouraged to fine-tune the model further for specific tasks or conversational use cases. The model is publicly available under the Microsoft Research License to support ongoing research in developing and aligning small language models.