BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:Grammar\, reasoning\, learning: Three short stories on comparative &\; rational analysis of language model capabilities - Dr. Andrew Lamp inen (Deepmind) DTSTART:20250306T160000Z DTEND:20250306T170000Z UID:TALK226258@talks.cam.ac.uk CONTACT:Shun Shao DESCRIPTION:Abstract:\nThere has been substantial debate about the capabil ities of language models—which aspects of language they can acquire\, wh ether they can be said to 'reason'\, and whether they can truly 'learn' in context. In this talk\, I will suggest that approaches from cognitive sci ence can provide useful tools for addressing these questions. Specifically \, I will focus on comparative methods (comparing capabilities across diff erent systems) and rational analysis (analyzing behavior as a rational ada ptation to an environment). I will illustrate different aspects of these i deas through three examples from our recent work: 1) comparing processing of recursive syntactic structures in language models and humans\, 2) evalu ating the way that both language models and humans entangle content in the ir responses to logical reasoning problems\, and 3) understanding how in-c ontext learning emerges from properties of training distributions. I will outline how these disparate phenomena can be understood using these cognit ive methods\, and the implications for evaluating and understanding langua ge model behaviors.\n\nBio:\nAndrew Lampinen's research bridges cognitive science and artificial intelligence\, often with a focus on how the comple x behaviors and representations of models\, agents\, or humans emerge from their learning experiences or data. His work covers topics ranging from i nterpretability\, to explanations as a learning signal\, to embodied intel ligence. He is currently a Staff Research Scientist at Google DeepMind. Be fore that\, he completed his PhD in cognitive psychology at Stanford Unive rsity\, and his BA in mathematics and physics at UC Berkeley. \n\n LOCATION:https://cam-ac-uk.zoom.us/j/97599459216?pwd=QTRsOWZCOXRTREVnbTJBd XVpOXFvdz09 END:VEVENT END:VCALENDAR