Teddi Worledge
PhD Candidate, Computer Science · Stanford University
About
Hi! I'm a computer science PhD student at Stanford, advised by Carlos Guestrin and supported by the NSF GRFP Fellowship. I often go by Teddi.
I care about making ML models reliable and trustworthy. I'm particularly interested in equipping users to interrogate model generations as model capabilities scale. Recently, I've been thinking about LLM citations, complementary human-AI performance, and what verifiability means to legal professionals!
Previously, I was at UC Berkeley, where I was fortunate to work with Ben Recht and Esther Rolf.
Recent Talks
- CodeX & Stanford Law School AI Initiative: CS x Law Evening · May 2026 — Given the reputational carnage from hallucinated legal AI citations, I'll be presenting on LLM citations and some predictions for the trajectory of legal AI.
- Civil Bench Seminar at the Los Angeles Superior Court · May 2026 — Shared our recent work on AI assistance for case review with the civil judges of LA Court––the largest trial court in the nation. Courthouses are racing to process crowded dockets and caseload is increasing even further due to readily available AI! We need to work on safely equipping courts with reliable assistive AI, while preserving judicial discretion.
- ACM CS & Law · March 2026 — Presented a law student user study of our AI case review assistant.
More Talks
- JURIX AI for Access to Justice · December 2025 — Presented on the development of our AI assistant for real-world case review.
- Code for Humanities Guest Lecture · June 2025 — Presented on LLMs to a group of Stanford humanities professors learning how to code! We covered different steps of the model training pipeline and why hallucinations are possible.
- Wonderfest · April 2024 — An open-to-all public session, where I got to talk about LLM citations and why they aren't reliable.
- NeurIPS ATTRIB · December 2023 — Workshop spotlight! Presented our unifying framework for training data attribution and LLM citations.
Papers
- AI Assistance for Court Review of Default Judgments Under Review, 2026
- The Extractive-Abstractive Spectrum: Uncovering Verifiability Trade-offs in LLM Generations Under Review at TACL, 2024
- Unifying Corroborative and Contributive Attributions in Large Language Models SaTML, 2024
- Representation matters: Assessing the importance of subgroup allocations in training data ICML, 2021
Gallery
The Starry Night & A Neural Algorithm of Artistic Style
Good Ol' Fashioned Sorting :)
