ChatYuan is a functional dialogue language model that supports Chinese and English bilinguals. ChatYuan-large-v2 uses the same technical solution as the v1 version, and has been optimized in terms of fine-tuning data, human feedback reinforcement learning, and chain of thought.

ChatYuan-large-v2 is one of the models in the ChatYuan series that achieve high-quality results with light weight. Users can perform inference on consumer-grade graphics cards, PCs and even mobile phones (INT4 only needs 400M minimum).

Based on the original functions of chatyuan-large-v1, v2 has been optimized as follows:

• Enhanced basic abilities. The original contextual question and answer and creative writing ability have been significantly improved.
• The ability to reject has been added. For some dangerous and harmful questions, I learned to refuse to answer them.
• Added Chinese-English bilingual dialogue ability.
• Added code generation functionality. Some optimizations have been made to the underlying code generation.
• Added table generation functionality. Make the generated table content and format more adaptable.
• Enhanced basic math skills.
• The maximum length is extended from 1024 tokens to 4096.
• Enhanced ability to simulate scenarios.