BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//events//Events//EN
CALSCALE:GREGORIAN
X-WR-CALNAME:Oct 22 - Advances in AI at Virginia Tech
REFRESH-INTERVAL;VALUE=DURATION:PT1H
X-PUBLISHED-TTL:PT1H
BEGIN:VEVENT
UID:event-rdvn9ich@ontown.app
DTSTAMP:20261008T195744Z
DTSTART:20261022T160000Z
SUMMARY:Oct 22 - Advances in AI at Virginia Tech
LOCATION:San Francisco\, United States
DESCRIPTION:AI Professionals San Francisco Machine Learning & Agentic AI\nJ
 oin our virtual meetup to hear talks from AI researchers at Virginia Tech!
 \n\nDate\, Time and Location\n\nOct 22\, 2026\n9:00 AM - 11:00 AM PST\nOnl
 ine. Register for the Zoom! (https://voxel51.com/events/advances-in-ai-at-
 virginia-tech-october-22-2026)\n\nMulti-Agent Communication: A framework\,
  diagnostic and mechanistic perspective\n\nMulti-agent LLM systems are inc
 reasingly used for collaborative reasoning\, debate\, and consensus\, yet 
 their communication dynamics remain poorly understood. This talk presents 
 a framework for studying multi-agent communication through diagnostic and 
 mechanistic perspectives.\n\nI will discuss CONSENSAGENT\, which improves 
 consensus by mitigating sycophancy\, alongside our diagnostic work on comm
 unication patterns and failure modes in real-world multi-agent debates. I 
 will then present ongoing work that moves toward a mechanistic understandi
 ng of how these interaction patterns arise internally\, with the broader g
 oal of making multi-agent systems more interpretable\, reliable\, and cont
 rollable.\n\nAbout the Speaker\n\nPriya Pitre (https://www.linkedin.com/in
 /priya-p-a31360117/) I am an Ph.D student in the Computer Science Departme
 nt at Virginia Tech (VT)\, co-advised by Dr. Xuan Wang and Dr. Naren Ramak
 rishnan.\n\nExposing and Improving Fine-Grained Visual Grounding Abilities
  of Lightweight Multimodal LLMs\n\nLightweight multimodal LLMs can localiz
 e whole objects effectively\, yet often struggle when a query targets a sm
 all object part or fine-grained visual detail. This talk presents a reason
 ing-guided framework that teaches compact models to ground parts through a
 n explicit coarse-to-fine process: first locating the parent object\, then
  identifying the requested part.\n\nA part-aware reinforcement-learning ob
 jective provides stage-wise rewards for object accuracy\, part containment
 \, and the consistency of the model’s self-critique. Using these techniq
 ues\, a compact 4B-parameter model achieves state-of-the-art zero-shot par
 t grounding while preserving its object-level performance.\n\nThese advanc
 es can be used to enable lightweight MLLMs to support detail-oriented task
 s in biology and robotics.\n\nAbout the Speaker\n\nKazi Mehrab  (https://w
 ww.linkedin.com/in/ksmehrab/)I a CS PhD candidate at Virginia Tech\, where
  I currently focus on multimodal LLMs and computer vision tasks\, includin
 g visual perception\, reasoning and grounding.\n\nUnderstanding Visual Gen
 erative Models for Precise Control\n\nDespite remarkable progress in image
  and video generation\, translating user intent into precise and consisten
 t visual outputs remains a challenge. This talk explores how understanding
  the representations within generative models can enable finer control ove
 r what they create.\n\nIt connects semantic image editing with composition
 al generation\, examining how visual concepts can be isolated\, manipulate
 d\, and combined while preserving their identity and surrounding content. 
 Building on these insights\, structured visual inputs provide a way to exp
 ress complex intent through subject references\, poses\, and spatial layou
 ts.\n\nThe discussion then extends from images to video\, where representa
 tions must evolve to preserve scene continuity while accommodating motion 
 and change. Together\, these directions establish a unified perspective on
  how visual representations can support controllable editing\, composition
 \, and coherent generation across space and time.\n\nAbout the Speaker\n\n
 Yusuf Dalva (https://www.linkedin.com/in/yusuf-dalva/) is a Ph.D. candidat
 e at Virginia Tech\, advised by Pinar Yanardag and affiliated with the San
 ghani Center for Artificial Intelligence and Data Analytics.
URL:https://ontown.app/e/rdvn9ich-oct-22-advances-in-ai-at-virginia-tech/
END:VEVENT
END:VCALENDAR