BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
X-WR-CALNAME:Physics Colloquium: Prof. Fan Yang
X-WR-TIMEZONE:Eastern Time (US & Canada)
BEGIN:VEVENT
DTSTAMP:20260712T183800Z
UID:tag:localist.com\,2008:EventInstance_49153067742281
DTSTART:20250402T200000Z
DTEND:20250402T210000Z
DESCRIPTION:Physics Colloquium: Towards Conceptual Understanding of Large L
 anguage Models with Prof. Fan Yang\, WFU Department of Computer Science. \
 n\nLarge Language Models (LLMs) have demonstrated impressive capabilities 
 in understanding and generating human language\, yet the internal mechanis
 ms by which they encode semantic knowledge and conceptual structures remai
 n largely opaque. In this talk\, I will introduce my recent exploration in
  probing LLM representations\, focusing on the depth-dependent emergence o
 f conceptual understanding and the structure of learned knowledge. The fir
 st part of the talk will introduce the concept of Concept Depth\, which po
 sits that simpler concepts are captured in shallower layers while more abs
 tract and complex ones require deeper processing. This framework is suppor
 ted by probing experiments across multiple LLM architectures\, shedding li
 ght on how different levels of semantic information are distributed within
  model layers. Building upon this\, I will discuss a complementary perspec
 tive that extends beyond single vector representations to Gaussian Concept
  Subspaces (GCS)\, a novel framework that models conceptual knowledge as a
  distribution rather than a single fixed direction in representation space
 . This approach improves robustness in identifying learned concepts and ha
 s practical implications for interpretability and intervention in model be
 havior. Finally\, I will present a vision for moving beyond empirical char
 acterization toward physical interpretations of LLMs\, drawing connections
  between deep learning representations and theoretical modeling of dynamic
 al systems. This future perspective invites interdisciplinary collaboratio
 n in physics\, neuroscience\, and AI\, aiming to ground the emergent behav
 iors of LLMs in a more principled framework.
GEO:36.1321;-80.27898
LOCATION:Olin Physical Laboratory\, 102
SUMMARY:Physics Colloquium: Prof. Fan Yang
URL;VALUE=URI:https://events.wfu.edu/event/physics-colloquium-towards-conce
 ptual-understanding-of-large-language-models-with-prof-fan-yang-wfu-depart
 ment-of-computer-science
CATEGORIES:Speakers
END:VEVENT
END:VCALENDAR