Atticus Geiger

Mixing Mechanisms: How Language Models Retrieve Bound Entities In-Context

Activation Steering via Contrastive Causal Mediation

Language Models Use Lookbacks to Track Beliefs

AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders

HyperDAS: Towards Automating Mechanistic Interpretability with Hypernetworks

MIB: A Mechanistic Interpretability Benchmark

How Do Transformers Learn Variable Binding in Symbolic Programs?