research Grounded Transitive RL Grounding divide-and-conquer value learning with temporal differences, so offline goal-conditioned RL stays correct when the dynamics are stochastic. Goal Representations & the State Pathway Handed a perfect goal representation, a goal-conditioned agent barely acts better. The headroom is in how it sees its own position. Language Guided Embodied Agents A vision–language model reflects on each failed episode in structured language, and the reflection becomes time-localised reward shaping for robot manipulation. Time Series Forecasting T3Time reads a series three ways (in time, in frequency, and as a language prompt) and lets the forecast horizon decide how to weigh them. Open World Amodal Counting CountOCC counts the objects you cannot see, by reconstructing their features under the occluder and checking its attention against an unoccluded view. Human Activity Detection U-ActionNet recognises human actions in drone footage. It crops each clip to the people in it, separates moving actors from the background in the frequency domain, and has a light version small enough for the edge. course Human–Robot Interaction
