BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Reading Between the Lines: Using Language Models to Amplify Human 
 Data in Robot Learning (MIT) - Andreea Bobu (MIT)
DTSTART:20251023T140000Z
DTEND:20251023T150000Z
UID:TALK237772@talks.cam.ac.uk
CONTACT:Lucas Resck
DESCRIPTION:Human-in-the-loop robot learning faces a fundamental data chal
 lenge that general machine learning doesn't: unlike settings where we can 
 collect massive offline datasets\, robots must learn from limited\, real-t
 ime human interactions. This creates a critical bottleneck: we need method
 s that can make the most of limited human input\, or\, in other words\, th
 at can learn a lot from a little. The key insight in this talk is that lar
 ge language models\, having been trained on vast amounts of human data\, a
 lready possess the common sense and semantic priors we need to fill in the
 se gaps. When someone demonstrates a task or gives feedback\, there's ofte
 n implicit information that seems obvious to humans but that robots overlo
 ok completely. I discuss three approaches that use language models to "rea
 d between the lines" of human input. I demonstrate how LLMs can take spars
 e human labels and enable robots to generalize to complex expressions\, ex
 tract hidden preferences that are implied by human behavior but not explic
 itly stated\, and identify missing task concepts based on the situational 
 context of human input. By strategically combining minimal human input wit
 h the rich prior knowledge embedded in language models\, we can achieve th
 e kind of sample-efficient learning that human-in-the-loop robotics demand
 s for real-world deployment.\n\nBio: Andreea Bobu is an Assistant Professo
 r at MIT in AeroAstro and CSAIL. She leads the Collaborative Learning and 
 Autonomy Research Lab (CLEAR Lab)\, where they develop autonomous agents t
 hat learn to do tasks for\, with\, and around people. Her goal is to ensur
 e that these agents' behavior is consistent with human expectations\, whet
 her they interact with expert designers or novice users. She obtained her 
 Ph.D. in Electrical Engineering and Computer Science at UC Berkeley with A
 nca Dragan in 2023. Prior to her Ph.D. she earned her Bachelor’s degree 
 in Computer Science and Engineering from MIT in 2017. She was the recipien
 t of the Apple AI/ML Ph.D. fellowship\, is a Rising Star in EECS and an R:
 SS and HRI Pioneer\, and has won best paper award at HRI 2020 and the Emer
 ging Research Award at the International Symposium on the Mathematics of N
 euroscience 2023. Before MIT\, she was also a Research Scientist at the AI
  Institute and an intern at NVIDIA in the Robotics Lab.
LOCATION:https://cam-ac-uk.zoom.us/j/97599459216?pwd=QTRsOWZCOXRTREVnbTJBd
 XVpOXFvdz09
END:VEVENT
END:VCALENDAR