← Search

Georgia Zhou

2 accepted papers

2025

Efficient Automated Circuit Discovery in Transformers using Contextual Decomposition

ICLR 2025poster

Automated mechanistic interpretation research has attracted great interest due to its potential to scale explanations of neural network internals to large models. Existing automated circuit discovery work relies on activation patching or its approximations to identify subgraphs in models for specifi…

Cited by 1SourcePDFScholar
2025

OMEGA: Can LLMs Reason Outside the Box in Math? Evaluating Exploratory, Compositional, and Transformative Generalization

NeurIPS 2025poster

Recent large language models (LLMs) with long-chain-of-thought reasoning—such as DeepSeek-R1—have achieved impressive results on Olympiad-level mathematics benchmarks. However, they often rely on a narrow set of strategies and struggle with problems that require a novel way of thinking. To systemati…

Cited by 0SourcecodeScholar