Special Sessions
Special SessionⅠ: Machine Reasoning
Session Chair: Prof. Ke Qin, University of Electronic Science and Technology of China, China
Session Co-chairs: Assoc. Researcher Shuang Liang, University of Electronic Science and Technology of China, China
Assoc. Researcher Dongyang Zhang, University of Electronic Science and Technology of China, China
Special Session Information:
With the explosive growth of Large Language Models (LLMs) and multimodal technologies, AI is evolving from perception toward cognition. However, the limitations of machine reasoning and the proliferation of generative content pose significant trust and security challenges for artificial intelligence. This session focuses on breakthroughs in LLM reasoning mechanisms, frontiers in multimodal fusion, and detection technologies for generated content. It aims to explore how to build next‑generation AI systems that are more reliable, secure, and capable of stronger reasoning, which is crucial for promoting the robust deployment of AI technologies. This session is intended for researchers, engineers, and graduate students from both academia and industry. Professionals working in natural language processing, computer vision, multimodal learning, AI safety and alignment, and related fields are particularly welcome. Participants will gain insights into the latest advances in machine reasoning and discuss core algorithms and defense strategies for generative content detection. Through cross‑disciplinary exchange, attendees will acquire new ideas for addressing complex reasoning and content security issues in real‑world applications and establish extensive international academic collaborations.
Below is an incomplete list of potential topics to be covered in the Special Session:
•LLM reasoning mechanisms and optimization (e.g., model design and training methods for logical reasoning, causal reasoning, and multi-step reasoning)
•Generative content detection and traceability technologies (e.g., identification algorithms and robustness evaluation for multimodal generative content such as text, images, audio, and video)
•Cutting-edge natural language processing technologies (e.g., knowledge graph, innovative methods and applications of semantic understanding, sentiment analysis, machine translation, and dialogue systems)
•Multimodal information processing and fusion (e.g., cross-modal representation learning, multi-source data alignment, modal conversion, and collaborative reasoning)
•Lightweight technologies for LLMs and multimodal models (e.g., model compression, quantization, pruning, distillation methods, and edge device adaptation)
•AI safety and alignment technologies (e.g., LLM hallucination mitigation, generative content risk prevention and control, and improvement of model interpretability and credibility)
•Practical application cases of related technologies (e.g., landing practices and effect analysis in fields such as intelligent healthcare, intelligent education, and enterprise services)
Special Session Keywords:
•LLM based Optimization
•Generative Content Detection
•Visual Question Answering
•Natural Language Processing
•Multimodal Information Processing
DDL: 2026-06-30