Dentsu's LLM Paper
2026-01-20 03:33:45

Dentsu Researcher's Paper on LLM Personalization Accepted at AAAI 2026 Workshop

A Major Leap for Dentsu in AI Personalization



Dentsu Research, headquartered in Minato, Tokyo, has reached a significant milestone with one of its researchers, Naoshi Osawa, whose single-authored paper on Large Language Models (LLMs) personalization has been accepted at the prestigious AAAI 2026 workshop entitled "Personalization in the Era of Large Foundation Models." The AAAI conference, organized by the Association for the Advancement of Artificial Intelligence, will be held from January 20 to 27, 2026, in Singapore. Dentsu is set to present a poster highlighting the accepted research during this highly esteemed international event.

Overview of the Research Paper



Osawa's research focuses on personalization in the context of generative AI. As generative AI continues its rapid advancement, it has found applications in various areas, including interactive systems and AI agents. However, many existing systems traditionally rely on a model of uniform responses, which leaves them ill-equipped to adapt to users' varied preferences and evolving intents.

In the accepted paper, Osawa proposes a novel method of expressing user preferences through two main elements: the direction of "likes and dislikes" and the intensity of these preferences. This innovative approach allows for a numerical integration of both past and current user preferences. Consequently, AI systems can now generate coherent responses that reflect real-time changes in users' preferences without requiring extensive retraining or significant system overhauls.

Notably, this proposed method is designed for ease of integration into existing AI services, making it accessible for companies looking to enhance their offerings. Through rigorous experimentation using multiple benchmark datasets, the study demonstrated the capability to generate responses that are more aligned with user preferences compared to traditional methods, ensuring stable and relevant engagement.

Author's Background: Naoshi Osawa



Naoshi Osawa is a key member of Dentsu's AI Transformation Center. Prior to joining Dentsu Research, he worked at a television station, focusing on broadcasting technology and the development of support systems for program production. At Dentsu, Osawa has centered his efforts on research and development in generative AI, playing a pivotal role in the creation of corporate generative AI solutions such as "Know Narrator." His contributions are instrumental in accelerating digital transformation (DX) and AI transformation strategies for businesses.

Future Directions for Dentsu Research



Dentsu Research is committed to leveraging the insights obtained from this research to further the development of advanced AI agent solutions and business support services. The firm aims to foster enhanced AI utilization within client companies, embarking on an exciting journey toward advancing AI capabilities.

About Dentsu Research



Dentsu Research aims to create a better society through technology, adhering to its corporate vision of "HUMANOLOGY for the future: building beyond today with people and technology." The organization synergizes its three core functions—system integration, consulting, and think tank activities—while engaging sincerely with society as a whole, including businesses, government agencies, and individuals. Dentsu continues to promote cross-innovation, breaking barriers across technology, industries, companies, and regions to inspire new value creation and a brighter future.

This ongoing commitment to innovation underlines Dentsu's aspiration for continuous evolution in a rapidly changing technology landscape. As AI continues to intertwine with daily life, Dentsu Research remains at the forefront, pioneering advancements that shape the future of human-tech interactions.


画像1

画像2

Topics Consumer Technology)

【About Using Articles】

You can freely use the title and article content by linking to the page where the article is posted.
※ Images cannot be used.

【About Links】

Links are free to use.