CHAI Seminar: Plenary Talk by Paulo Shakarian, KG Tan Endowed Professor of Artificial Intelligence, Syracuse University
DATE: Monday, September 21, 2026, 12:00-1:00 PM
SPEAKER: Dr. Paulo Shakarian, KG Tan Endowed Professor of Artificial Intelligence, Syracuse University
TITLE: Improving LLM Performance with Metacognition
IN PERSON: Simone Center 1600 Atrium
ABSTRACT: Recent advances in large language models (LLM's) have primarily stemmed from the so-called scaling laws whereby improvement increases with large networks and more data. Correctness in this paradigm is defined in terms of model fit to historical data. However, the use of LLM's for general use-cases causes an objective function mismatch: the LLM was trained with one objective but is being applied to a problem with a different one. LLM's can fail for several reasons: they may not be employed in a constrained manner, they may provide results inconsistent with domain knowledge, and they may exhibit common failure modes in certain use cases. We present three approaches that can alleviate these problems. First, we study the use of abductive inference to constrain the use of the LLM for targeted text reframing. Then, we show how domain knowledge can inform an LLM cybersecurity agent by enforcing logical constraints. Third, we examine how the use of psychology-inspired metacognitive techniques can reduce errors in a small language model, allowing for local queries and limiting use of more expensive LLMs.
BIO: Paulo Shakarian is the K.G. Tan Endowed Professor of Artificial Intelligence at Syracuse University where he directs the Leibniz Lab. Shakarian has made notable contributions in the areas of logic programming, neurosymbolic AI, security, and data mining. His academic accomplishments include four best-paper awards, over 100 peer-reviewed articles, 12 issued patents, and 8 published books. Shakarian was recently named to the Defense Science Study Group, previously held fellowships with DARPA and the New America Foundation, and has led many grants and contracts from major funders including Amazon, DARPA, and many defense-related funders. Shakarian previously held positions with West Point and Arizona State (where he earned tenure), created and led an AI-cyber startup company, and previously served in the U.S. Army. He holds a Ph.D. in computer science from the University of Maryland, College Park.
Event Snapshot
When and Where
Who
Open to the Public
| Cost | FREE |
Interpreter Requested?
No