Mitigating Sycophancy in Large Language Models through Truth-Aware Evaluation and Penalty-Based Feedback
Large Language Models (LLMs) can sometimes agree with users even when their claims are factually incorrect, a behavior known as sycophancy. This project investigates how user opinions, misleading claims, and expressed confidence influence LLM responses. Students will evaluate LLMs using factual questions presented under different user viewpoints and develop a framework to measure behaviors such as unsupported agreement, answer switching, and resistance to misleading information. The project will also explore penalty-based feedback and prompting strategies to reduce sycophantic behavior while maintaining factual correctness and response quality.
Requirements: Python programming; basic knowledge of Machine Learning, Data Mining, or Data Science; familiarity with Natural Language Processing is preferred
Faculty: Iqra Ameer, Humaira Farid, Vinayak Elangovan
Probing the Roles of ILF3 and RNP Complexes in Axonogenesis
This project will investigate what factors affect complex formation between ribonucleoproteins and interleukin enhancer binding factor 3 (ILF3). This may play an important role in global gene expression, immune response, and cell development. Cell culture and biochemical assays may be used to assess this. I already selected my students for the 2026-2027 academic year.
Requirements: Minimum 3.0 GPA, passed BIOL 230W and CHEM 112/113/210 with a C or better. Preferred: Passed CHEM 213 with a C or better.
Faculty: Shana Wagner
Effects of Organic Compounds on Neuronal Health
Ever wonder how your thoughts turn into movements? Part of the process involves neurons sending chemical messengers from an axon to a dendrite in a delicate network called the nervous system. Neurons can be damaged by age, neurodegenerative diseases, and cancerous tumors. This project aims to test whether two series of organic chemicals (thiazolidinones and γ-butyrolactones) with specialized functional groups on conserved scaffolds display neuroprotective or chemotherapeutic properties using a neuroblastoma cell line. Cell culture and molecular biology assays will be used to obtain data. I have already selected students for the 2026-2027 academic year.
Requirements: Minimum 3.0 GPA, passed BIOL 230W and CHEM 112/210 with a C or better. Preferred: Passed CHEM 213 with a C or better.
Faculty: Shana Wagner
Transforming Landscapes with Undergraduate Community-Engaged Research and the Commonwealth Arboreta Network
Now that the Abington campus has attained level one Arboretum Accreditation, this project entails continuing to make changes and maintaining the changes already made on campus that reflect the University Sustainable Operations Council's goal of more biodiverse landscaping.
Requirements: include an interest in environmental and sustainability sciences, willingness to apply the scientific method to research questions, and the ability to plant trees and work in the garden.
Faculty: Michele Grinar
Pulsar Investigations using Radio Astronomy
Students will investigate the properties of pulsars. Pulsars are rapidly spinning, tiny stars which are the remnants of supernova explosions. Students will remotely use radio telescopes at the Green Bank Observatory.
Requirements: Students should have completed a course in introductory physics, be able to use algebra and trig, and be able to use EXCEL or code in Python.
Faculty: Ann Schmiedekamp ([email protected]), Carl Schmiedekamp ([email protected])
Potash Production in Colonial New England and in the New Republic
Potash was an early cash crop for colonists, and citizens of northern Atlantic states used it to make soap, glass, and other products. This project will research the production and analysis of potash from wood ash to determine factors that affected purity and yield.
Requirements: An interest in chemistry and/or history. Chemical techniques such as extraction and titration will be used; completion of Gen Chem 1 lab (CHEM 111) is recommended.
Faculty: Kevin Cannon
Infectious Disease Predictor and Visualizer
This project is an infectious-disease spread simulator and herd-immunity visualizer. It models how pathogens travel through a population using a grid-based Susceptible-Infect-Resolved (SIR) model. The simulation is currently driven by real-world data; disease R0 values, death rates per thousand, and vaccination rates at the state level across the United States. The user selects a disease and a geographic granularity (county, state, or country), and the program runs the simulation, animates the spread of the disease, scales results to real population sizes, and displays interactive maps of the outcomes. This research effort will expand the model simulations, adding various efficacy rules for each disease, improving the user experience, researching better sources of complete infectious disease data, and incorporating worldwide data for other countries.
Requirements: Students are required to have a GPA > 3.0 and strong knowledge of Python (CMPSC 131 & 132) or beyond. Experience with other high-level programming languages (Java, C++, etc.) is beneficial. Ability to work independently and good communication skills are required.
Faculty: Mark Grande
Design Methods for Traditional Wood Joinery
Traditional wood joinery methods use fitted geometric shapes to connect members as opposed to using fasteners. The goal of this investigation is to determine if current wood connection design engineering methods are appropriate for predicting the capacity of traditional joinery.
Requirements: Students intending to major in engineering.
Faculty: David Brown
Probing the Roles of ILF3 and RNP Complexes in Axonogenesis
This project will investigate what factors affect complex formation between ribonucleoproteins and interleukin enhancer binding factor 3 (ILF3). This may play an important role in global gene expression, immune response, and cell development. Cell culture, electrophoresis, immunoprecipitation, western blotting, and RNA isolation may be used to assess this.
Requirements: Minimum 3.0 GPA, passed BIOL 230W and CHEM 112/113/210 with a C or better. Preferred: Passed CHEM 213 with a C or better.
Faculty: Shana Wagner
Infectious Disease Predictor and Visualizer
This project is a grid-based SIR (Susceptible-Infect-Resolved) disease spread simulator and herd-immunity visualizer. It uses real U.S. state-level data—including R0 values, death rates, and vaccination rates—to animate pathogen spread at the county, state, or country level, scaling results to real populations and displaying interactive outcome maps. Future work will expand the model with disease-specific efficacy rules, enhance user experience, find better infectious-disease data sources, and add worldwide data for other countries.
Requirements: Students are required to have a GPA > 3.0, strong knowledge of Python (CMPSC 131 & 132) or beyond. Experience with other high-level programming languages (java, C++, etc.) is beneficial. Ability to work independently and good communication skills are required.
Faculty: Mark Grande
Design Methods for Traditional Wood Joinery
Traditional wood joinery methods use fitted geometric shapes to connect members as opposed to using fasteners. The goal of this investigation is to determine if current wood connection design engineering methods are appropriate for predicting the capacity of traditional joinery.
Requirements: Students intending to major in engineering.
Faculty: David Brown
Concrete for Martian exploration and environmental sustainability
Fabrication, testing, and data analysis of special types of concrete (and similar materials). Of primary focus this year are in-situ material for exploration of Mars and environmentally-friendly concrete to be used on our planet.
Requirements: GPA > 3.6
Faculty: Masataka Okutsu
3D Scanning–Robotic Arm Integration for Intelligent Systems
This project introduces students to the fundamentals of combining 3D scanning technologies with robotic arm systems for automation and precision tasks. Participants will gain hands-on experience in capturing 3D models and programming robotic movements to interact with scanned objects.
Requirements: Advanced programming skills, Advanced skills in hardware design and circuits, GPA > 3.2, self-motivated, and can work independently.
Faculty: Yi Yang
Bird Window Collisions on the Penn State Abington Campus
This project will investigate the incidences of bird collisions with windows on the Penn State Abington campus. The aim of the project is to estimate the number of bird collisions with windows and campus locations that are at an increased risk of collisions.
Requirements: Science majors with an interest in ecology, conservation, or birds and a minimum GPA of 3.0.
Faculty: Les Murray
Harnessing Large Language Models for Intelligent Computing Systems
This project aims to exploit Large Language Models (LLMs) and natural language processing (NLP) techniques to enhance the quality and efficiency of various data-driven tasks. The key sub-projects include 1. Enhancing Sentiment Analysis: Using LLMs to capture nuanced emotional and contextual signals in textual data. 2. Improving Time Series Analysis: Applying LLMs to extract patterns, forecast trends, and interpret temporal data. 3. Providing Objective Interpretations from Data: Generating unbiased summaries and insights from structured and unstructured datasets. 4. Enabling Automation: Utilizing LLMs to perform routine tasks, decision-making, and workflow orchestration. 5. Supporting Root Cause Analysis (RCA): Integrating LLMs with conventional RCA frameworks to identify underlying causes of anomalies or failures through intelligent reasoning and contextual understanding.
Requirements: Experience in Tensorflow or PyTorch, experience in data preprocessing, strong self-motivation, and commitment to the project.
Faculty: Janghoon Yang
Controlling a Robot Arm using a Large Language Model (LLM)
The objective of this research project is to develop a controller for a low-cost robot arm using a Large Language Model (LLM). Combining LLM abilities for visual object detection along with code generation, an agent-based system is under development to control a low-cost robot arm using natural language commands (ex: “pick up the blue block”, “align the blocks in a horizontal line”). This technology can impact areas such as intelligent manufacturing, robot surgery, 3D printing, assembly/disassembly, construction, art, circuit fabrication, home robotics, healthcare, etc.
Requirements: Students are required to have a GPA > 3.0, strong knowledge of Python (CMPSC 131 & 132) or beyond, and experience with LLM API programming is beneficial. Ability to work independently and good communication skills are required.
Faculty: Robert Avanzato
Effects of derivatized organic compounds on gastrointestinal stromal tumor cells
Cancer remains one of the deadliest diseases in humans. One often understudied form of this disease is gastrointestinal stromal cancer. Previous studies have shown that thiazolidinones can selectively inhibit the growth of cancer cells in culture. We will test the effects of differentially modified thiazolidinones, synthesized by our organic chemists, on the growth and reproduction of two gastrointestinal stromal tumor cells. These molecules have different halogens attached to the at either the meta or para position. This project will allow students to learn techniques in cell culture, microscopy, and cellular quantification.
Requirements: Students are typically selected from the biology or integrative science majors. No prior research experience is required. Students working with me usually have a GPA of at least 3.0. I have already selected my students for the 2024-25 academic year.
Faculty: Eric Ingersoll [email protected]
Underwater Object Detection using Sonar and Deep Learning (AI)
The purpose of the research project is to apply deep learning (AI) techniques to detect and identify important artifacts (shipwrecks, partially buried man-made structures, pipes, animal life, etc.) in side-scan sonar data collected by an underwater robot. Deep learning is a subset of machine learning and artificial intelligence. Deep learning uses a convoluted neural network (CNN) to identify patterns and objects in images (as well as other data, such sound and text). Deep learning has proven very successful in many application areas as tumor detection in x-ray images and CT scans, natural language detection, and face recognition.
Requirements: Advanced programming skills in Python or MATLAB, knowledge of computer vision and AI (preferred), GPA 3.0 or above.
Faculty: Robert Avanzato ([email protected])
AI and Image Segmentation: Advancing SAM for Enhanced Stereo Video and Point Cloud Processing
This project investigates using AI-driven image segmentation, specifically the Segment Anything Model (SAM2), to enhance stereo video and point cloud processing. By integrating SAM2 with 3D data techniques, the research aims to improve segmentation accuracy and efficiency, addressing challenges in depth estimation and object recognition.
Requirements:
- Advanced programming skills
- Advanced skills in hardware design and circuits
- GPA > 3.2
- Self motivated and can work independently
Faculty: Yi Yang ([email protected])
Classification of Japanese Papers Based on Deep Learning of Optical Coherence Tomography Images
This project focuses on classifying Japanese paper types using deep learning techniques applied to Optical Coherence Tomography (OCT) images. By leveraging advanced image analysis, the research aims to accurately differentiate between various types of traditional Japanese papers, which are often challenging to classify due to subtle structural differences. The project will develop and train deep learning models to identify unique characteristics in OCT images, contributing to the preservation and study of cultural heritage materials.
Requirements:
- Advanced programming skills
- Advanced skills in hardware design and circuits
- GPA > 3.2
- Self motivated and can work independently
Faculty: Yi Yang ([email protected])
Transforming Landscapes with Undergraduate Community Engaged Research and the Commonwealth Arboreta Network
As Penn State Abington prepares for construction on a new Academic Building, this project entails making small changes on campus that reflect the University Sustainable Operations Council's goal of more biodiverse landscaping. Throughout fall of 2024 and spring of 2025, students will conduct research to identify locations, species, and numbers of trees, shrubs, and supportive ecosystems to be planted on campus with the goal of acquiring level one Arboretum Accreditation.
Requirements: None.
Faculty: Michele Grinar ([email protected])
Smart Electric Wheelchair
This project aims to build a smart electric wheelchair using an H-Bridge motor controller for efficient drive control. It enables smooth, precise movement, including forward, reverse, and turning. Smart features like obstacle detection and assistive technology integration will enhance safety and user independence.
Requirements: Advanced pogramming skills
Faculty: Vinayak Elangovan ([email protected]) Co-advisor: Yi Yang ([email protected])