AI BriefingUpstage Ranks No. 1 on Hugging Face LLM Leaderboard
Key point
Upstage's 70B model took first place on the Hugging Face LLM Leaderboard.
Details
Upstage's newly released 70B generative AI model scored 72.3 points on Hugging Face's Open LLM Leaderboard, surpassing GPT-3.5 (71.9 points) to take the No. 1 spot.
Just before this, Upstage's 30B model had achieved an average of 67 points, surpassing LLaMA 2 70B to become the first Korean LLM to reach No. 1. This time, through fine-tuning based on more data and the latest LLaMA 2 70B, Upstage pushed its ranking even higher, cementing its position at the top globally.
Drawing on its prompt engineering and fine-tuning capabilities built up through Kaggle award-winning experience, building KLUE, winning in 4 ICDAR categories, and operating AskUp, Upstage mobilized all of its expertise to create a model surpassing GPT-3.5 in just about two months. Going forward, the company plans to commercialize this technology to target the Private AI market, which trains models solely on a company's internal data, emphasizing that it can be adopted at lower cost than large-scale LLMs while reducing security and hallucination issues.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.