Elliot Pickens

About Me

I'm a PhD student in the Computer and Information Science department at the University of Pennsylvania, where I am co-advised by Duncan Watts and Chris Callison-Burch. Prior to joining the CIS department, I was briefly a student at the University of Wisconsin-Madison, and before that, I was a pre-doctoral research fellow at MSR-NYC, where I was advised by David Rothschild.

I earned my Master's in Operations Research and Information Engineering (ORIE) at Cornell Tech and studied mathematics as an undergrad at Carleton College.

My research is currently centered around language models. I am researching how they work and how to use them effectively.

This page serves as an approximate catologue of my ongoing research, tutorials, and other writing.

Content & Writeups

Works in Progress

• An LLM Political Bias Benchmark - derived from real world human-LLM conversations about political topics.
• Approaches for understanding how training data influences LLM outputs - using tools from mechanistic interpretability, representation learning, optimization, and Bayesian statistics
• Exploring how LLMs bridge stylistics and semantics - through a mechanistic lens
• Improving uncertainty quantification in mechanistic interpretability
• Using tabular foundation models - for neural posterior estimation, agentic workflows, and other things

Publications, Articles & Other Research (Note: This probably is not up to date)

• Pre-trained Tabular Foundation Models as Versatile Summary Networks for Neural Posterior Estimation
• Framing in the Presence of Supporting Data: A Case Study in U.S. Economic News
• The DataSquad Experiment: Lessons for Preparing Data and Computer Scientists for Work
• Warped Front Pages
• Optimization Under Uncertainty: The Scenario Approach and Guaranteed Error Machines--> >
  → GitHub Repository
• Graphical Inference with Convolutional Neural Networks
  → USRESP Award Link
k

Probability & Statistics Series

• Expectation Pt. 4 (Moments) - PDF
• Expectation Pt. 3 (Properties of Expectation) - PDF
• Expectation Pt. 2 (Variance) - PDF
• Expectation Pt. 1 (Intro) - PDF
• Random Variables Pt. 4 (Markov Chains) - PDF
• Random Variables Pt. 3 (Functions of a Random Variable) - PDF
• Random Variables Pt. 2 (Multivariate Distributions) - PDF
• Random Variables Pt. 1 (The Basics) - PDF
• Intro to Bayes' Theorem - PDF
• Basics of Conditional Probability - PDF

Project Notes & Explorations

• Dynamic Pricing in a Competitive Environment - PDF
• Natural Language Image Search - PDF
• Modeling Citi Bike Availability - PDF
• A Brief Intro to Bayesian Belief Networks
  → GitHub Repository
• Some Notes on Reinforcement Learning - GitHub
• Hidden Markov Models (DRAFT) - PDF
• Replicating Bayesian Distributional Word Embeddings - PDF
• Paraphrase Detection
  → GitHub Repository
• Quick Analysis of Metro Transit Dwell Times - HTML Slides
• Quantifying Smart Growth (Case Studies: Wellington & Colorado Springs) - PDF

Learning Materials & Notes

• Graph Analysis and Random Graphs - PDF
• Online Algorithms: Ski Rental Problem - PDF

Contact Information

Email: epickens@seas.upenn.edu

GitHub: github.com/epickens