BLOOMZ 1.7B

About

BLOOMZ 1.7B is a multilingual large language model designed for zero-shot learning, enabling it to follow instructions in various languages without prior training 147. As part of the BLOOMZ and mT0 family, it builds upon the BLOOM and mT5 models. Featuring a decoder-only transformer architecture, it was fine-tuned on the xP3 dataset, which includes diverse tasks and languages 147. While effective in translation, text generation, and question answering, its performance heavily relies on clear prompts and sufficient context 6. However, it is unsuitable for high-stakes applications due to the potential for generating inaccurate information and inherent biases in its training data 3.

Capabilities