Information

Mercari Selected for GENIAC Robotics Foundation Model R&D Project, Implemented by METI and NEDO

2026.09.09
#PhysicalAI #Robotics

Mercari, Inc. (“Mercari”) is pleased to announce that its research project “Research and Development on Leveraging AI Robots for Listing and Inspection Tasks” has been selected for the Generative AI Accelerator Challenge (GENIAC), a national initiative spearheaded by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO). The aim of the national initiative is to strengthen the development capabilities and social implementation of generative AI in Japan.

Mercari is in the midst of driving its transformation into an “AI-Native Company,” focusing on rebuilding all of its organizations, development processes, and products around AI technologies. For this project, Mercari’s R&D organization R4D will conduct research and development on physical AI to automate Mercari’s inspection and listing tasks. R4D will use Mercari’s inspection and shipping facility, which handles a wide variety of items, as a real-world testing environment. Researchers will work on three key initiatives: building on Vision-Language-Action (VLA) models1 to create multimodal models that integrate touch and force sensing; developing camera-aware VLA models that can handle viewpoint changes; and building a data collection framework that captures and transfers the actual sensations of human work. Using this technology, researchers will strive to optimize logistics efficiency in Mercari’s crossborder transactions.

This is the second time this year that a Mercari project has been selected for GENIAC. The first project, announced in June, was titled “Development of a High-Precision Search and Recommendation Platform Model for the Secondary-Distribution Market Using Generative Retrieval Technology.”2

1. AI model that interprets visual input captured from a camera and natural language commands given by a human to generate the actions of a robot in a physical space.
2. Mercari Selected for GENIAC—Generative AI Accelerator Challenge Led by METI and NEDO (https://about.mercari.com/en/press/news/articles/20260604_geniac/ )

The challenges of automating the inspection tasks supporting the expansion of crossborder transactions and of working with current physical AI models

In September 2025, Mercari launched the Mercari Global App, a worldwide app that allows buyers overseas to more easily, safely, and securely buy from Mercari users in Japan. The app is currently available in Taiwan, Hong Kong, and the US, with Mercari working to expand its reach by aiming to launch the same product to over 100 additional countries and territories over the mid to long term.

One of the critical pieces of infrastructure behind this crossborder growth is Mercari’s inspection and shipping facility. To facilitate safe and secure transactions, Japan-sourced items purchased by users overseas are collected at Mercari’s facility, where they undergo inspection before they are shipped overseas. Because Mercari’s crossborder transactions deal with items across various categories, items vary widely in shape and condition. Inspection also involves handling deformable items such as fabrics and bags, and manipulating components like buttons and zippers. These are tasks that are difficult to automate with conventional model-based robot control alone. ⁠⁠

In recent years, VLA models—which generate actions directly from vision and language input—have emerged and are expected to be useful in environments like these. However, in adopting current VLA models for use in item inspection and other such tasks, the following three major pain points arise:

  • The robot’s behavior relies solely on visual information input through cameras, and it cannot gauge the appropriate amount of force to apply to an item based on its feel and response
  • If the camera angle during inference differs even slightly from the images the model was trained on, the trained robot becomes unable to perform the appropriate operations
  • Data collection through robot teleoperation is inefficient

In this project, Mercari’s inspection and shipping facility—which handles a wide variety of irregular items—will serve as a real-world testing environment for technical innovations aimed at overcoming these limitations of VLA models. The success of this project will do more than just further accelerate the rapidly expanding growth of Mercari’s crossborder business. It aims to contribute to the establishment of physical AI infrastructure that can be applied to the physical tasks of various industries.

About GENIAC

The Generative AI Accelerator Challenge (GENIAC) is a project spearheaded by METI and NEDO with the aim of strengthening the development capabilities and social implementation of generative AI in Japan.

For more details on this research, see the accompanying article on the R4D website.

https://r4d.mercari.com/en/news/20260909_geniac_physicalai_details

Mercari will continue to connect communities across industry, academia, and government, contributing to the development of cutting-edge technology and social implementation while pioneering the path toward undiscovered value.