Open-source Large Language Model Framework
8.3k 2026-04-18
LianjiaTech/BELLE
BELLE is an open-source project dedicated to fostering the development of Chinese conversational large language models, aiming to make LLMs accessible to everyone.
Core Features
Provides open-source Chinese conversational LLM models.
Offers instruction training data, code, and application scenarios.
Optimized for Chinese language, fine-tuned with ChatGPT-generated data.
Includes advanced speech recognition and multimodal LLM capabilities.
Supports various training methods like RLHF (PPO, DPO) and efficient inference.
Detailed Introduction
BELLE (Be Everyone's Large Language model Engine) is an initiative to democratize large language models, particularly for the Chinese language. Instead of focusing on pre-training, BELLE emphasizes instruction tuning on top of existing open-source LLMs, providing high-quality training data, models, and code. The project aims to significantly lower the barrier for research and application of Chinese LLMs, continuously evaluating different data and algorithms to improve model performance and accessibility for a wide range of users and developers.