Riiid's AI model ranks #1 in HuggingFace LLM leaderboard

SAN RAMON, Calif., Oct. 9, 2023 /PRNewswire/ -- Riiid, a leading provider of AI-powered education solutions, is pleased to announce that its latest generative AI model was ranked number one on HuggingFace, a leading platform for building and sharing machine learning models, 

The HuggingFace Open LLM (Large Language Model) Leaderboard ranks the performance of more than 500 open-source generative AI models worldwide. Amid fierce global competition, dozens of new LLMs are released every day and when participating companies or individuals submit their LLMs to HuggingFace, the models are evaluated and the Leaderboard is updated. 

Riiid's LLM, submitted in September, scored 73.69 points and was ranked second in the world. Riiid's latest model, 'Sheep-duck-llama-2,' submitted in October, scored 74.07 points and was ranked first. 

Sheep-duck-llama-2 is a fine-tuned model from llama-2-70b, and is used for text generation. It is built on the HuggingFace Transformers library and uses datasets in the Orca-style and Alpaca-style formats. The model has been evaluated on several metrics, including ARC (25-shot), HellaSwag (10-shot), MMLU (5-shot), and TruthfulQA (0-shot).

"As an AI technology company that is not limited to the education field, this first-place ranking proves that our LLM learning capabilities and tuning technology are at the forefront of the world," said Youngjun Jang, CEO of Riiid. "We expect that our prompt engineering, refined through the operation of AI for education and learning, will generate good synergies with our fine-tuning technology."

ABOUT RIIID

Riiid is working to democratize quality education globally. It has partnered with diverse education companies and schools to provide personalized learning experiences to students in more than 10 countries. Riiid most recently signed an MOU to provide AI technology for public education in Brazil.  Riiid researchers continue to develop novel architectures and ever-higher-performing AI models, verifying the learning impact with real-world data and publishing papers at international AI conferences.

For more information, contact Heejung Shin at [email protected]

SOURCE Riiid