In this presentation, Yuhui will present a paper on building generalist robot policies using a vision-language foundation model. The authors propose a flow matching architecture built on top of a pre-trained vision-language model (VLM) to inherit Internet-scale semantic knowledge. Trained on a large, varied dataset, the model can follow language instructions, perform tasks in a zero-shot setting, and learn new skills via fine-tuning.

In this presentation, Manisha will present a paper which explores how computational cognitive science helps us understand human decision-making using tools like probability theory, reinforcement learning, and statistical modeling. It reviews models that explain both how people make decisions (forward models) and how they reason about others' decisions (inverse models). The authors highlight recent progress in integrating black-box learning with theory-driven approaches and reframe heuristics as rational strategies under cognitive constraints. The work bridges cognitive science with control and optimization perspectives.

In this talk, Joshua will discuss a paper that reviews recent progress in physically assistive robots developed to help individuals with disabilities perform everyday tasks such as moving, eating, and personal care. As these robots become safer, more capable, and affordable, real-world deployment is increasing. The paper highlights major research trends in interaction methods, autonomy, and adaptability, and outlines frameworks and future directions for the field.

In this talk, Matias will explore a paper that introduces Bayesian Object Models (BOMs), a framework enabling robots to build rich, uncertainty-aware models of unseen objects from limited interaction. These models capture both structural and dynamic object properties using a differentiable probabilistic program. By combining a tree structure sampler with a physics engine, BOMs enable efficient gradient-based Bayesian inference, outperforming recent neural and physics-based alternatives.

In this talk, Emanuele will discuss about a class of TAMP problems and survey algorithms for solving them, characterizing the solution methods in terms of their strategies for solving the continuous-space subproblems and their techniques for integrating the discrete and continuous components of the search.

What do we mean when we say we need to design assistive agents that help us in our daily activities? In this talk, Rim will aim to answer this question by surveying several seminal works and recent advances in designing assistive agents, stemming from decision-making and information theory.

Sal introduces an AI-assisted bronchoscope robot that enables novice doctors to perform safe and expert-level lung examinations, addressing global disparities in access to skilled bronchoscopic care.