Reverse-engineering specific algorithms implemented by transformer models, such as induction heads, indirect object identification, and greater-than comparisons. Focuses on finding interpretable computational circuits within neural network activations.

Explores how biological systems minimize prediction error through active inference. Investigates the brain as a hierarchical generative model that actively samples the world to confirm predictions and minimize surprise.

Investigating capabilities that appear suddenly at specific model scales, including multi-step reasoning, instruction following, and few-shot learning. Questions whether these are genuine phase transitions or measurement artifacts.

Implementation and optimization of surface code error correction schemes for stabilizing logical qubits. Focuses on achieving fault-tolerant quantum computation through topological codes with local check operators.

Development of ionizable lipid nanoparticles (LNPs) that enable efficient cellular uptake and endosomal escape of mRNA therapeutics. Critical breakthrough enabling COVID-19 vaccines and next-generation gene therapies.

Testing whether cosmic acceleration arises from modifications to general relativity rather than a cosmological constant. Includes f(R) gravity, galileons, and massive gravity theories with observational constraints from structure formation.

Developing practical methods to compute Φ (phi) from neural recordings and test Integrated Information Theory predictions. Addresses computational intractability through approximations and identifies empirical signatures of consciousness.

Deep learning approaches to predict protein 3D structure from sequence alone, achieving experimental accuracy. Explores attention mechanisms, multiple sequence alignments, and iterative refinement for structural biology applications.