AI summary
Written by AI from the official notes. Check them for exact details.The deepseek-chat model has been upgraded to DeepSeek-V2-0517, enhancing instruction following and JSON output accuracy.
- Improved instruction following accuracy from 63.9% to 77.6%.
- Optimized system instruction handling for better user experience.
- JSON parsing rate increased from 78% to 97%.
Why it matters: Users relying on accurate instructions and JSON outputs should consider upgrading for improved performance.
Full release notes
The deepseek-chat model has been upgraded to DeepSeek-V2-0517. The model has seen a significant improvement in following instructions, with the IFEval Benchmark Prompt-Level accuracy jumping from 63.9% to 77.6%. Additionally, on API end, we have optimized model ability to follow instruction filled in the system" part. This optimization has significantly elevated the user experience across a variety of tasks, including immersive translation, Retrieval-Augmented Generation (RAG), and more.
The model's accuracy in outputting JSON format has been enhanced. In our internal test set, the JSON parsing rate increased from 78% to 85%. By introducing appropriate regular expressions, the JSON parsing rate was further improved to 97%.