DeepSeek upgrades deepseek-chat to DeepSeek-V2-0517, greatly improving instruction following and JSON output
Original title:deepseek-chat
AI summary
The deepseek-chat model was upgraded to DeepSeek-V2-0517, with IFEval Benchmark Prompt-Level accuracy jumping from 63.9% to 77.6%, and instruction following in the API's "system" area was also optimized, enhancing experiences in tasks such as immersive translation and RAG. JSON format output accuracy improved, with the parsing rate on the internal test set rising from 78% to 85%, and further to 97% after introducing appropriate regular expressions.
Key Takeaways
- The deepseek-chat model was upgraded to DeepSeek-V2-0517,with IFEval Benchmark Prompt-Level…
- 6%,and instruction following in the API's "system" area was also optimized,enhancing experiences…
- JSON format output accuracy improved,with the parsing rate on the internal test set rising from…
Quick Facts
- Published
- Updated
- Source
- DeepSeek:API 更新日志
- Category
- Model Releases
- AI score
- 68
DeepSeek V2 的指令跟随从及格线拉到优秀线,IFEval 提升近 14 个点,沉浸式翻译和 RAG 体验会有明显改善,做应用层的值得试一下。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: DeepSeek upgrades deepseek-chat to DeepSeek-V2-0517, greatly improving instruction following and JSON output. May 17, 2024. https://new.nat6.net/en/news/cms8ndnte01q3rof1c1jcxw2g/