AI for Science Agentic laboratories and automated research systems, from scientific world models to benchmarks for drug discovery. Benchmarks and Evals Benchmarks, environments, and evaluation methodology for coding, tool use, computer use, and expert-level professional work. Enterprise Agents Qualifying agents for real business workflows on reliability, human oversight, and operating cost, not just benchmark accuracy. Agentic RL Training Training recipes, reward hacking, data mixing Agent Teams Coordination, communication, and cost efficiency in multi-agent systems, so that teams of agents can take on long-horizon work that no single agent finishes reliably. MLSys Orchestration, parallelism, kernel-level efficiency, and the machinery that moves models from training runs into products. Probabilistic Graphical Models Computer Vision Video retrieval at TRECVID, from shot boundary detection and concept detection to search. Self-driving Vehicles Mining vehicle behaviors from multi-sensor data for scenario parsing, surveillance, and collision warning.