NAVER : Team NAVER Takes Global AI Lead with 347 Papers Published and Over 30,000 Citations in 5 Years

- Awarded "Best Paper" at key AI conferences this year, Team NAVER excels in AI safety research

- As a global AI industry leader, Team NAVER demonstrates proficiency in pilot studies, hyperscale AI development, and business success

- Future plans to elevate AI safety research and establish global leadership in "responsible AI technology"

December 20, 2023

NAVER CLOUD and NAVER have had 347 AI research papers accepted at top-tier global conferences from 2019 to 2023. Starting with 29 papers in 2019 and reaching approximately 100 papers in each of the past two years^*, the increasing trend in publication volume underscores Team NAVER's swift advancement in AI technological capabilities.

*The number of papers by NAVER CLOUD and NAVER published at top-tier global AI conferences (2019-2023) - 2019: 29 papers, 2020: 46 papers, 2021: 72 papers, 2022: 107 papers, 2023: 93 papers

The influence of each research piece by NAVER CLOUD and NAVER is evident, garnering over 30,000 citations, with more than 10,000 in 2023 alone, as per Google Scholar. Their impact in the AI research domain is notable, reflected in a 2022 ranking^* that places them sixth, ahead of Google and Intel, among AI companies, based on the percentage of top 100 papers in research influence.

*This ranking is based on the analysis by the global AI research platform Zeta Alpha (March 2023)

In particular, Team NAVER excelled in AI safety research, an area gaining increasing importance, securing the technological leadership to build safe AI. Their achievement includes a research proposal^* for a tool for detecting personal information leaks in hyperscale language models, recognized in the top 10% at the Neural Information Processing Systems (NeurIPS) 2023. This research was a collaborative effort between NAVER CLOUD and the University of Tübingen-NAVER Joint Research Center for reliable AI. Since 2021, NAVER and the University of Tübingen have been working together on researching·developing methods to reduce AI biases and enhance interpretability. In addition, NAVER's collaboration with Seoul National University (SNU) at the SNU-NAVER Hyperscale AI Research Center focuses on bolstering the trustworthy hyperscale AI, with their joint research^** on improving hyperscale language model trustworthiness being accepted at the prestigious Association for Computational Linguistics (ACL) 2023 Conference.

*ProPILE: Probing Privacy Leakage in Large Language Models (https://arxiv.org/abs/2307.01881) **Query-Efficient Black-Box Red Teaming via Bayesian Optimization (https://arxiv.org/abs/2305.17444)

Research on "Sustainable and Responsible AI" also gained noteworthy recognition. A study^* showcasing the societal benefits of hyperscale AI via NAVER's AI-based elderly care service, "CLOVA CareCall," earned recognition as a "Best Paper", which is the top 1% research at CHI 2023 - the premier conference in the field of human-computer interaction. CLOVA CareCall was also featured in an Asia-Pacific Economic Cooperation (APEC) report^** published in November as an exemplary case for strengthening the 'Bio-Circular Green Economy model'. Moreover, research^*** aimed at developing a dataset to mitigate hyperscale AI's biased responses in contentious social issues received a "Best Paper" nomination at the ACL 2023 conference.

*Understanding the Benefits and Challenges of Deploying Conversational AI Leveraging Large Language Models for Public Health Intervention (https://dl.acm.org/doi/10.1145/3544548.3581503) **APEC Case Study: Technology Empowering BCG Economy Model (November 2023) ***SQuARe: A Large-Scale Dataset of Sensitive Questions and Acceptable Responses Created Through Human-Machine Collaboration (https://arxiv.org/abs/2305.17696)

In a year where AI has taken center stage for tech companies, these accomplishments establish Team NAVER as a formidable player among global tech giants beyond the nation's highest level. In August, NAVER unveiled "HyperCLOVA X," an upgraded version of its self-developed hyperscale AI, the third of its kind globally, expanding its market to include both individual and corporate clients. As the only Korean company, NAVER participated in the organizing committee of NeurIPS and the International Conference on Machine Learning (ICML) and was officially invited to the UK government's AI Safety Summit in November as one of two Korean firms, where it engaged in discussions with top AI firms like OpenAI, DeepMind, and Anthropic on developing safe AI.

Ha Jung-woo, Executive Director of AI Innovation Center at NAVER CLOUD, stated, "It is common to encounter international researchers who read all papers published by NAVER in the global academic community. Like this NAVER's research is highly esteemed, evident in the extensive citations and conference acceptances. As the significance of safe AI continues to grow in everyday life and the workplace, NAVER is committed to intensifying its focus on AI safety research and leading the global technology initiative."