The Food Hazard Detection Challenge, Task 9 of SemEval 2025, has successfully concluded! This competition gathered AI researchers and food safety experts to develop machine learning models that detect food hazards in real-world reports.
About the Challenge
SemEval 2025 Task 9: The Food Hazard Detection Challenge focused on evaluating explainable classification systems for food-incident report titles collected from the web. These algorithms have the potential to enhance automated crawlers, improving food safety monitoring from sources like social media. Given the high economic impact of food hazards, transparency in detection methods was a key focus.
The challenge consisted of two main tasks:
- Text Classification for Food Hazard Prediction (ST1): Predicting the type of hazard (e.g., Salmonella, pesticides) and the affected product category (e.g., dairy, meat).
- Food Hazard and Product Detection (ST2): Identifying the specific hazard and product mentioned in food safety reports.
Participants worked with a dataset of 6,644 reports, covering 1,142 products and 128 different hazards. The models were evaluated based on the macro F1-score, ensuring fairness across hazard types and product categories.
Key Insights
The challenge attracted strong participation, with top-performing teams employing advanced AI techniques such as Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG) to achieve high accuracy. These innovations pave the way for AI-driven food safety monitoring, helping detect risks before they escalate into public health threats.
What's Next?
Following the competition, participants will document their methods and findings in system description papers, which will be submitted to SemEval 2025, co-hosted with ACL 2025 in Vienna. These papers will contribute valuable insights to the AI research community, advancing the field of explainable AI in food safety.
EFRA will continue working towards an AI-driven food risk prevention system, leveraging the knowledge gained from this challenge. Future developments will focus on refining AI models, improving data aggregation methods, and ensuring greater transparency in food safety monitoring.
Read the full article to explore the competition details, methodologies, and what’s next for AI in food safety.
