Tec-Do leads MARS2 workshop at ECCV 2026 with US$100,000 prize pool

ECCV 2026, one of the world's premier conferences in computer vision and machine learning, ran from September 8 to 12 in Malmö, Sweden, managed by the European Computer Vision Association (ECVA) and sponsored by Google, Meta, Apple and Amazon. Within that setting, Tec-Do led the second MARS2 Workshop on September 9, standing out as the only Agentic Commerce-focused workshop at the conference to be led by a Chinese technology company. The organizing committee included researchers from Tsinghua University, the University of Oxford, Nanyang Technological University and Seoul National University, and invited speakers included Paul Pu Liang of MIT, Yarin Gal of the University of Oxford and Shanxin Yuan of Queen Mary University of London, who spoke on multimodal reasoning, long-chain reasoning, zero-shot generalization and agentic systems.

The workshop's theme framed the shift from perception-oriented 'fast thinking' to deliberate 'slow thinking' as especially important for Agentic Commerce. As AI agents take on a greater role in commercial decision-making, the argument goes, they must do more than understand content: they must identify relevant evidence, reason across multiple modalities and explain how decisions are made. To test those abilities, MARS2 hosted a multimodal reasoning challenge with a total prize pool of US$100,000 across three tracks — Multimodal Advertisement Comprehension, Video Temporal Grounding, and Marketing Strategy Decoding and Conversion Analysis — evaluating holistic understanding, evidence localization and higher-order reasoning. The competition drew 64 teams and more than 1,060 submissions, including researchers from the University of Science and Technology of China, Nankai University and Sun Yat-sen University, as well as technical teams from ByteDance, JD.com and Xiaohongshu (RedNote), with award-winning teams presenting their approaches during the workshop.

How much of this translates into practice is likely to hinge on whether the deliberate reasoning the workshop advocates can be shown to work in real commercial settings, not just benchmark tasks. For the technical teams and academic groups that took part, the value may lie less in the prize pool than in the shared benchmark and the winning methods now on display — which is where the next round of testing is likely to happen.