Max Tegmark

A Benchmark for Vericoding: Formally Verified Program Synthesis

Scaling Laws For Scalable Oversight

Dense SAE Latents Are Features, Not Bugs

KAN: Kolmogorov–Arnold Networks

Low-Rank Adapting Models for Sparse Autoencoders

Efficient Dictionary Learning with Switch Sparse Autoencoders

Not All Language Model Features Are One-Dimensionally Linear

Are Sparse Autoencoders Useful? A Case Study in Sparse Probing

On the creation of narrow AI: hierarchy and nonlocality of neural network skills

Remarkable Robustness of LLMs: Stages of Inference?

Interpretable Patterns in Random Initialization Unveil Final Representation

Decomposing The Dark Matter of Sparse Autoencoders

Generalization from Starvation: Hints of Universality in LLM Knowledge Graph Learning

The Geometry of Concepts: Sparse Autoencoder Feature Structure