TLDR: Bloomberg’s AI engineers have made significant contributions to the field of information retrieval, presenting their latest research at SIGIR 2025, the premier international forum for the domain. Their work includes a notable paper on image-text retrieval robustness and their involvement in a workshop on financial information retrieval in the age of generative AI.
Bloomberg’s commitment to advancing artificial intelligence in financial technology is once again evident as its AI engineers present groundbreaking research at the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2025), held from July 13-18, 2025, in Padua, Italy. This prestigious event serves as a critical platform for showcasing new research results and innovative systems in the broad field of information retrieval (IR).
Among the notable contributions from Bloomberg AI is the paper titled “Benchmark Granularity and Model Robustness for Image-Text Retrieval: A Reproducibility Study.” This research, co-authored by Mariya Hendriksen (who interned at Bloomberg AI) and Shuo Zhang (a Senior Research Scientist at Bloomberg AI in London), along with Ridho Reinanda, Mohamed Yahya, Edgar Meij, and Maarten de Rijke, delves into crucial aspects of image-text retrieval (ITR). The paper explores the challenges of evaluating ITR models, particularly focusing on the granularity of benchmarks and the robustness of models against various input perturbations. This work is vital for developing more reliable and accurate multimodal information access systems in real-world applications, where noisy, ambiguous, and perturbed inputs are common. The official code for this paper is also publicly available, promoting reproducibility and further research in the community.
Beyond individual papers, Bloomberg AI is actively shaping the discourse in specialized IR domains. Shuo Zhang, a Senior Research Scientist at Bloomberg AI, announced that their workshop, “FinIR: The 2nd Workshop on Financial Information Retrieval in the Era of Generative AI,” has been accepted at SIGIR 2025. This workshop underscores Bloomberg’s leadership in applying cutting-edge AI, including generative AI, to the complex and data-rich financial sector, addressing unique challenges in financial information retrieval.
Bloomberg’s AI department, comprising over 350 AI practitioners, is at the forefront of building sophisticated products and features that leverage novel innovations in AI. The company is heavily investing in AI to enhance search, discovery, and workflow solutions, utilizing advanced technologies such as transformers, gradient-boosted decision trees, large language models (LLMs), and dense vector databases. Their initiatives include developing and deploying robust Retrieval-Augmented Generation (RAG) systems, curating high-quality data for model training, and building evaluation frameworks for continuous improvement. Bloomberg’s AI engineers are also designing and implementing tools for LLM-powered search agents to handle complex client queries, thereby shaping the company’s generative AI ecosystem and scaling these solutions for thousands of users.
Also Read:
- AI Innovations Surge: Grok 4 Redefines Performance Benchmarks
- AI-Powered Image Generation: A Lucrative Investment Frontier
Since 2009, Bloomberg has been developing AI applications to process and organize the ever-increasing volume of structured and unstructured financial information, uncovering signals and producing analytics across all asset classes. Their active participation and publication at SIGIR 2025 highlight their ongoing commitment to pushing the boundaries of information retrieval research and its practical applications in the global capital markets.
