AIA Forecaster: Technical Report
Nov 2025
An LLM-based forecasting system combining intelligent news source searches, supervisor agents, and statistical calibration. Achieves expert-level performance on ForecastBench.
Papers
Inertial Navigation System Drift Reduction Using Scientific Machine Learning
Dec 2024
Master's thesis on reducing inertial navigation system drift using physics-informed neural ODEs. Cut drift by over 60% in GPS-denied settings.
Low Complexity Solutions For Interpolating Deep Neural Networks
Dec 2024
Extends double descent research to deep neural networks with a novel algorithm for learning low-complexity models.
How Do Transformers 'Do' Math? A Study of Meaningful Intermediates for Linear Regression
Dec 2024
Investigating how transformers perform mathematical operations through analysis of intermediate representations in linear regression tasks.
Bridging the Gap: Integrating Hebbian Plasticity into Deep Learning
Dec 2024
Exploring the integration of Hebbian learning principles into modern deep learning architectures using Marr's computational framework.
Efficiency of Class Naming at MIT—Numbers or Names?
Mar 2024
Modeling the brevity-disambiguation trade-off in MIT's course lexicon with implications for LLM prompts and retrieval systems.
Evaluating GPT-3 in Simulated Cybersecurity Scenarios
May 2023
Evaluating the performance of GPT-3 in simulated cybersecurity scenarios, focusing on graph-based problems.