BlueLM 7B

About

BlueLM 7B is an advanced open-source language model developed by vivo AI Lab, equipped with 7 billion parameters and based on the GPT-Neo architecture. It supports both base and chat functionalities, capable of handling up to 32K token contexts, which is advantageous for tasks requiring complex text comprehension. Trained on 2.6 trillion tokens from primarily Chinese and English sources, with additional Japanese and Korean text, it excels in tasks such as text generation, summarization, and question-answering, demonstrating strong performance in benchmarks like C-Eval and CMMLU. Though open for research and commercial use, users should be mindful of potential biases and the lack of specific safety fine-tuning.

Capabilities