UPSTAGE-QANDA-KT DEVELOPS A MATH-SPECIFIC LANGUAGE MODEL THAT SURPASSES CHATGPT
2024/1/8
UPSTAGE-QANDA DEVELOPS OPEN AI AND MATH GPT THAT SURPASSES MS AS PART OF KT STRATEGIC PARTNERSHIP
GPT-4 surpasses MS ToRA 13B performance... Achieves first place among models under 13B size
CONQUER THE DIFFICULT AREA OF LLM MATHEMATICS THROUGH ORGANIC COOPERATION BETWEEN UPSTAGE, QANDA, AND KT
(Upstage = 2024/1/8) Upstage (CEO Seong-Hoon Kim) announced that 'MathGPT (tentative name)', a mathematics domain specialized program jointly developed with Masspresso (CEO Lee Yong-jae), the operator of the AI-based learning platform 'Qanda', and KT, is part of Open AI. It was announced on the 8th that it had achieved a new world record, beating ChatGPT and Microsoft (MS) models.
Last November, Upstage and Qanda began developing MathGPT as part of a strategic partnership with KT. Upstage learned Qanda's high-quality mathematical data and fine-tuned the natural language-based language model to solve complex mathematical problems through logical reasoning and programming.
Accordingly, the two companies developed MathGPT with the 13B parameter and achieved an encouraging report card, surpassing the best-in-class MS 'ToRA 13B' in a representative benchmark test that evaluates the mathematical abilities of language models such as 'MATH' and 'GSM8K'. This is a new record, ranking first for models under 13B.
ACHIEVING THE HIGHEST PERFORMANCE SIMULTANEOUSLY IN TWO MEASURES, THE MATH BENCHMARK, WHICH CONSISTS OF 12,500 DIFFICULT MATH PROBLEMS, AND THE GSM8K BENCHMARK, WHICH TESTS ARITHMETIC OPERATIONS WITH 8,500 ELEMENTARY SCHOOL MATH PROBLEMS, IS AN UNPRECEDENTED ACHIEVEMENT IN RECENT TIMES IN THE WORLD. ALL.
In particular, MathGPT surpassed ChatGPT performance in the benchmark test average and even surpassed GPT-4 in the MATH benchmark. In the difficult mathematical domain, a domestically developed small-sized model has set a meaningful mark by beating big tech companies such as OpenAI and MS.
THIS ACHIEVEMENT WAS LARGELY DUE TO THE ORGANIC COOPERATION BETWEEN UPSTAGE, QANDA, AND KT. THANKS TO KT'S FULL INFRASTRUCTURE SUPPORT THROUGH STRATEGIC PARTNERSHIPS, QANDA'S UNRIVALED MATH-SPECIFIC KNOW-HOW, HIGH-QUALITY LEARNING DATA, AND ITS OWN LANGUAGE MODEL 'SOLAR' HAVE SURPASSED GLOBAL BIG TECH TO CREATE THE WORLD'S NO. 1 LLM. UPSTAGE'S MODELING TECHNOLOGY COMBINED TO PRODUCE THE BEST RESULTS.
Upstage, which has confirmed visible results in the education field through MathGPT, will move forward to reorganize the LLM market with Solar at the forefront. We plan to focus on strengthening our global structure as a stepping stone for full-fledged overseas expansion beyond the domestic market, covering various industries such as finance, distribution, medical care, and entertainment.
KIM SEONG-HOON, CEO OF UPSTAGE, SAID, “IT IS MEANINGFUL TO DEVELOP A WORLD-CLASS MATH-SPECIFIC LANGUAGE MODEL THAT WILL SURPASS CHATGPT THROUGH COOPERATION WITH QANDA AND KT.” HE ADDED, “IN THE FUTURE, UPSTAGE WILL WORK IN VARIOUS AREAS BASED ON THE WORLD’S NO. 1 LLM TECHNOLOGY.” “WE WILL LEAD GENERATIVE AI INNOVATION,” HE SAID.
LEE YONG-JAE, CEO OF QANDA, SAID, “THE TECHNOLOGY AND KNOW-HOW OF THE THREE COMPANIES, QANDA, UPSTAGE, AND KT, WERE COMBINED TO CREATE MEANINGFUL RESULTS THAT SURPASSED THOSE OF GLOBAL BIG TECH.” HE ADDED, “THROUGH QANDA’S AI TUTOR, WE ARE PROVIDING EDUCATIONAL EXPERIENCES TO STUDENTS NOT ONLY IN ASIA BUT AROUND THE WORLD. “WE WILL INNOVATE,” HE SAID.
-
Geun-Kyo Kim | Brand Communication General Director | keunkyo@upstage.ai
Seongbeom Bae | Brand Communication Manager | sungbae@upstage.aiDownload press release
-
Upstage is a leading domestic AI startup established in October 2020. Upstage stands out in the large language model (LLM) industry by taking first place on the Hugging Face leaderboard with performance exceeding ChatGPT's benchmark score for the first time in OpenLLM history. Based on these technologies, we present a reliable private LLM standard that maximizes data security and solves hallucination, helping companies conveniently use cutting-edge technology. In addition, Upstage's Chat AI 'AskUp' has over 1.4 million users, establishing itself as the largest AI service in Korea. Document AI Pack, another Upstage representative solution, utilizes AI OCR technology that has won the world's most prestigious OCR competition to automate documents by increasing efficiency and accuracy. By optimizing document processing through a pre-trained model with minimal data, cost and time are dramatically minimized compared to manual methods. Lastly, through the education program 'EduStage', we are also actively engaged in the educational content business that fosters differentiated professional talent who can be immediately put into AI business through hands-on education that incorporates AI business experience and solid AI basic education.
Upstage is comprised of members from global big tech companies such as Google, Apple, Amazon, NVIDIA, Meta, and Naver, and has participated in many world-renowned AI academic societies such as NeurlPS, ICLR, CVPR, ECCV, WWW, CHI, WSDM, and DMLR. We are solidifying our unrivaled leadership in AI technology by publishing excellent papers and becoming the only domestic company to win double-digit gold medals in the online AI competition Kaggle. While working as a professor at the Hong Kong University of Science and Technology, Upstage CEO Kim Seong-hoon won the ACM Sigsoft Distinguished Paper Award, the best paper award, four times for his research on bug prediction and automatic source code generation that combined software engineering and machine learning, and won 10 awards at the International Conference on Software Maintenance. He is considered a world-class AI guru who received the most influential paper award in 2018, and is also widely known as an instructor of 'Deep Learning for Everyone' with a total of more than 7 million views. Additionally, Upstage's co-founders include CTO Lee Tal-seok, who led Naver Visual AI/OCR and achieved world-class results, and CSO Park Eun-jung, who led the model team of Papago, the world's best translator.