Приём заявок на воркшоп CVPR 2026: Безопасность мультимодальных агентов

The artificial intelligence world is developing rapidly, and with it grows the need to ensure the safety and reliability of the systems being created. This question is particularly acute when it comes to multimodal large language models, which are becoming increasingly powerful and autonomous. Within the framework of the CVPR 2026 conference, the sixth AdvML@CV workshop will take place, entirely devoted to this timely topic.

Multimodal large language models (LLM) are systems capable of processing and generating information presented in various formats: text, images, audio, and video. Integrated into agents, they can make decisions and act in the real world, which opens enormous possibilities, but also creates new risks. Imagine an LLM-agent managing a complex production line or an autonomous vehicle. In case of malfunction or deliberate attack, the consequences could be catastrophic.

The AdvML@CV workshop aims to bring together researchers and practitioners working in the field of machine learning security to exchange experience and develop strategies to protect against potential threats. Special attention will be paid to the following issues: attacks on multimodal LLMs, methods for detecting and preventing such attacks, increasing model resilience to adversarial effects, developing secure architectures and protocols.

Participation in the AdvML@CV workshop is an opportunity to learn about the latest achievements in multimodal LLM security, share your developments, and establish contacts with leading experts. Call for papers is already open, and the organizers invite all interested parties to present their research and developments.

The safety of multimodal LLMs is not just an academic task, but an urgent necessity. The future of many industries and the well-being of society as a whole depend on how successfully we can protect these systems from malicious actors and ensure their reliable operation. The AdvML@CV workshop is an important platform for discussing these issues and developing joint solutions. Participation in it will not only expand your knowledge but also contribute to creating a safer and more reliable world of artificial intelligence.

In conclusion, the AdvML@CV workshop at CVPR 2026 is a key event for everyone interested in the safety of multimodal large language models. It is an opportunity to learn about the latest research, share experience, and contribute to ensuring the reliability of these powerful technologies. Do not miss the chance to be part of this important conversation.