Accepted Competitions

---

Management

The competition organizers will oversee all aspects of the competition, including its management. For details on participation and updates, please refer to the competition's official website.

Conference’s Role

The conference committee will not participate in the management of the competition or the selection of winners.

Publications

The competition organizers may recommend selecting competition-related papers for inclusion in the conference's companion proceedings. Authors of accepted papers are required to attend the conference in person to present their work. The authors will be required to register for the published papers which is similar to the condition of the workshop papers.

In-Person Workshop

Some organizers may attend the conference to host a workshop showcasing the winners' solutions and discussing the competition outcomes. Further updates regarding this decision will be provided later.

1. Multimodal Dialogue System Intent Recognition Challenge 

• Background: The Internet has become a primary communication channel for delivering customer service. A key challenge in web-based customer support is the efficient recognition of multimodal intentions within service dialogues. By leveraging advanced AI and large language model (LLM) technologies, accurately identifying customer intentions across multiple interactions can greatly enhance the customer experience and service quality, while also supporting business intelligence initiatives for companies. This competition, co-organized by TaoTian Group, WWW2025, and Tianchi, provides a dataset comprising text and images across over 40 consumer image scenarios and intent classification tasks. The goal is to push forward intent recognition technology in multimodal dialogue scenarios within the e-commerce sector. 
• Organizers: TaoTian Group 
• Hosting organizations: Customer Operations Department of TaoTian Group, Zhejiang Lab, Alibaba Cloud Tianchi 
• Competition website: https://tianchi.aliyun.com/competition/entrance/532277?lang=en-us
• In-person Workshop: YES

2. Google - Unlock Global Communication with Gemma

• Abstract: With over 7,000 languages and countless cultural differences, AI has the potential to foster global understanding. In a step towards broader linguistic inclusion, we're launching a Kaggle competition focused on adapting Gemma 2, Google's open model family, for 73 eligible languages. These languages were selected to represent a diverse range and to align with the expertise of our judging panel for effective evaluation. Our initial focus on these languages will allow us to establish a robust foundation of techniques and resources that will later enable us to support under-resourced languages.
• Organizers: Google
• Competition website: https://www.kaggle.com/competitions/gemma-language-tuning
• In-person workshop: No

3. Brick-by-Brick: Cyber-Physical Building Data Classification Challenge

• Abstract: Buildings account for nearly one-third of global energy consumption and carbon emissions, making their optimization essential in combating climate change. Cyber-Physical Buildings, enabled by the integration of Internet-of-Things (IoT) devices and advanced data analytics tools like AI, offer a smart and effective approach to energy management. A key challenge, however, lies in automating the semantic labeling of IoT devices to ensure machine-interpretable data. The "Brick-by-Brick: Cyber-Physical Building Data Classification Challenge" aims to tackle this challenge by classifying time-series data from IoT devices within buildings. Participants will engage with a dataset consisting of over 10,000 time-series streams collected over three years across three buildings, representing 91 unique semantic classes.
• Organizers: UNSW, Australia
• Hosting organisations: UNSW, and CSIRO
• Competition website: https://www.aicrowd.com/challenges/brick-by-brick-2024
• In-person workshop: TBD

4. AgentSociety Challenge: Designing LLM Agents for User Modeling and Recommendation

• Abstract: Online review platforms play a pivotal role in shaping social interactions and information exchange in the digital age, providing spaces for people to share experiences and seek insights. The AgentSociety Challenge aims to explore the potential of large language model (LLM) agents in understanding user behavior and enhancing user experiences within these platforms. Participants are invited to compete in two tracks: the User Modeling Track, which focuses on modeling user preferences and simulating online reviews, and the Recommendation Track, which emphasizes delivering personalized recommendations across diverse scenarios. The Challenge utilizes large-scale public datasets from prominent online review platforms, complemented by a textual simulator to provide interactive feedback for LLM agents. By encouraging participants to design LLM agents within a modular design space, the Challenge seeks to advance the integration of LLM technologies into real-world web applications. Winning modular designs will be reusable and adaptable, offering long-term value to academic research, industry applications, and the broader Web community.
• Organisers: Tsinghua University & Infinigence AI
• Competition Website: https://tsinghua-fib-lab.github.io/AgentSocietyChallenge/
• In-person workshop: YES

5. SmartMem: Memory Failure Prediction for Reliable Cloud Service

• Abstract: Uncorrectable Errors (UEs) of Dual RandomAccess Memory(DRAM) have been identified as a major failure cause in data centers and the multi-bit UE failure of High Bandwidth Memory (HBM), which severely threatens the availability and reliability of cloud service, entire computing clusters, and web-based applications. Forecasting UEs before enacting preemptive maintenance measures has emerged as a viable strategy for diminishing server outages and some machine-learning-based solutions have also been proposed. However, the UEs prediction presents several challenges: data noise and extreme imbalance as the UEs are exceedingly rare in memory events; heterogeneous data sources as the DRAMs in the field come from different manufacturing or architecture platforms; distribution shifts due to the hardware aging; and latent factors due to the dynamic access mechanism. These challenges become even more pronounced in web-centric contexts, where memory usage patterns can vary significantly across applications such as content delivery networks, e-commerce platforms, and real-time analytics. We cure a real-world memory error dataset that contains both micro and bit information and present a two-stage challenge for a more efficient and generalized event prediction solution. We believe the competition will provide a breeding ground to foster discussions and further progress on several important research topics towards real-world ML applications in web-based and cloud environments.
• Organisers: Huawei
• Hosting organisations: Huawei, Technical University of Berlin, USTC, University of Coimbra
• Competition Website: TBD
• In-person workshop: TBD