Marc Anton Finzi

Predicting the Performance of Black-box Language Models with Follow-up Queries

Antidistillation Sampling

Compute-Optimal LLMs Provably Generalize Better with Scale

Eliciting Black-Box Representations from LLMs through Self-Queries

The Promises and Pitfalls of Language Models for Structured Numerical Data

Non-Vacuous Generalization Bounds for Large Language Models

Searching for Efficient Linear Layers over a Continuous Space of Structured Matrices

Unlocking Tokens as Data Points for Generalization Bounds on Larger Language Models

The No Free Lunch Theorem, Kolmogorov Complexity, and the Role of Inductive Biases in Machine Learning

Diffusing Differentiable Representations