A challenge on causal reasoning and scenario retrieval for autonomous driving. Task, data, and baselines are available on the challenge page; winners will be announced during the workshop.
From Modalities to Propositions: A Language-Centric Framework for Multimodal Intelligence. Image, video, and text are represented as bags of atomic propositions for interpretable reasoning, retrieval, and data curation.
Read paper →CASCADE: A Spatio-Temporal-Causal Reasoning Representation and Dataset for Driving. 2,066 human-annotated driving clips with 34K+ elements for machine-verifiable reasoning.
Read paper →Advancing Explicit Behavior Modeling with Stochastic Iterative Scoring for end-to-end driving.
Read paper →World models and synthetic data generation to scale training and evaluation for autonomous driving and Physical AI.
Current openings on the PAIS team. New positions will be listed here as they open.
Research, prototype, and develop synthetic data, generative world models, and simulation techniques to improve the training, evaluation, generalization, and safety of next-generation autonomous driving systems.
