James Campbell

I am an ML PhD student at CMU.

Currently, some of my interests include: LLM agents, inference-time compute, interpretability, and alignment.

I am very eager to meet other ambitious people! If you'd like to chat, please don't hesitate to get in touch.

James Campbell

About

I am an ML PhD student at Carnegie Mellon University advised by Yiming Yang. At a high level, my research aims to understand the ingredients for general intelligence while ensuring its safety.

As an undergrad at Cornell, I worked on LLM interpretability and truthfulness, and was a primary contributor to the papers Representation Engineering and Localizing Lying in Llama. I also worked at Gray Swan AI on adversarial robustness and evaluations. Previously, I did work in computational neuroscience, physics, and deep learning theory. I am also the creator of ProctorAI, AI-Timeline.org, and co-author of AidanBench.

Papers

AidanBench

AidanBench: Stress-Testing Language Model Creativity on Open-Ended Questions

Aidan McLaughlin*, James Campbell*, Anuja Uppuluri*, Yiming Yang

Presented at NeurIPS 2024 Language Gamification Workshop

Summary: We benchmark the ability of models to come up with as many original answers as they can to open-ended questions.

Localizing Lying in Llama

Localizing Lying in Llama: Understanding Instructed Dishonesty on True-False Questions

James Campbell*, Phillip Guo*, Richard Ren*

Presented at NeurIPS 2023 SoLaR Workshop

Summary: We prompt Llama-2-70B-chat to lie and localize mechanisms involved using activation patching and linear probing.

Representation Engineering

Representation Engineering: A Top-Down Approach to AI Transparency

Andy Zou, Long Phan*, Sarah Chen*, James Campbell*, Phillip Guo*, Richard Ren*, et al.

ArXiv Preprint

Summary: We introduce the field of Representation Engineering which seeks to understand and control LLM's using a top-down approach.

Biological Plausibility

Considerations of Biological Plausibility in Deep Learning

James Campbell

Published front page of the Cornell Undergraduate Research Journal (CURJ)
Winner of the $300 James E. Rice Award

A literature review of learning algorithms and their biological plausibility

Projects

AI-Timeline

AI-Timeline.org

AI-Timeline.org is an interactive timeline built in React documenting the last 10 years of AI progress.

ProctorAI

ProctorAI

Proctor is a multimodal AI companion that watches your screen and yells at you if it sees you being unproductive. It has over 300 stars on Github.

Theoretical Bound

Theoretical Bound on the Weights of a Neural Network under Nesterov's Accelerated Gradient Flow

I solved an "open question" posed in a previous paper by deriving this bound on the weights of a neural network. This was done in summer of 2021 during an REU at Johns Hopkins under Rene Vidal.

Neural Network Linearization

Neural Network Linearization for Out-of-Distribution Robustness

In summer of 2022, I worked with Yi Ma on improving out-of-distribution robustness using the empirical neural tangent kernel.

Data-driven clustering

Data-driven clustering of neural responses to a large set of natural images

I investigated the clusterability of visual representations in the brain. This work was done in the Computational Connectomics Lab at Cornell and was presented as a poster at OHBM.

Additive Compositionality

Exploring Additive Compositionality in the Brain

I conducted experiments attempting to understand the extent to which semantic representations in the brain exhibit additive compositionality, i.e. does representation(A) + representation(B) = representation(A+B).

Kaggle Competition

Cornell Machine Learning Kaggle Competition

I came in first out of 155 participants in a Kaggle competition hosted by Cornell's big machine learning class.

CereBERTo

CereBERTo: Improving Distributional Robustness with Brain-Like Language Representations

Improving the out-of-distribution robustness of BERT and GPT-2 by pretraining them to predict brain fMRI data.

music

Command Line Music Player

A Spotify-like music player that operates from the command line. Gained over 140 stars on Github.

Contact

One of my great pleasures is meeting interesting and ambitious people! Please don't hesitate to reach out. I am often in Pittsburgh, New York, SF, and Berkeley.