Search for: All records

Large language models (LLMs) exhibit a wide range of promising capabilities -- from step-by-step planning to commonsense reasoning -- that may provide utility for robots, but remain prone to confidently hallucinated predictions. In this work, we present KnowNo, which is a framework for measuring and aligning the uncertainty of LLM-based planners such that they know when they don't know and ask for help when needed. KnowNo builds on the theory of conformal prediction to provide statistical guarantees on task completion while minimizing human help in complex multi-step planning settings. Experiments across a variety of simulated and real robot setups that involve tasks with different modes of ambiguity (e.g., from spatial to numeric uncertainties, from human preferences to Winograd schemas) show that KnowNo performs favorably over modern baselines (which may involve ensembles or extensive prompt tuning) in terms of improving efficiency and autonomy, while providing formal assurances. KnowNo can be used with LLMs out of the box without model-finetuning, and suggests a promising lightweight approach to modeling uncertainty that can complement and scale with the growing capabilities of foundation models. 

With their ability to embody users in physically distant spaces, telepresence robots have gained popularity in environments including hospitals, schools, and offices. However, with platforms lacking in individuation and social presence, users often personalize telepresence robots with clothing and accessories to increase their recognizability and sense of embodiment. Toward understanding personalization preferences, as well as perceptions of personalized platforms, we conducted a series of five studies that investigate patterns in personalization of a telepresence robot and evaluate the impacts of common personalizations along five dimensions (robot uniqueness, humanness, pleasantness/unpleasantness, and people's willingness to interact with it). Finding a strong preference for the use of clothing and headwear in Studies 1-2 (N=52), we systematically manipulated a robot's appearance using these items and evaluated the qualitative and quantitative impacts on observer perceptions in Studies 3-4 (N=160). Observing that personalization increased perceptions of uniqueness and humanness, but also decreased positive responding, we then investigated the associations between personalization preferences and perceptions via a fifth study (N=100). Across the five studies, tensions emerged between operators' interest in using wigs and interlocutors' dislike of wigs. This result highlights a need to consider both operator and interlocutor perspectives when personalizing telepresence robots.