Neel Nanda

Narrow Finetuning Leaves Clearly Readable Traces in Activation Differences

Steering Evaluation-Aware Language Models To Act Like They Are Deployed

Thought Branches: Interpreting LLM Reasoning Requires Resampling

Emergent Misalignment is Easy, Narrow Misalignment is Hard

What's the plan? Metrics for implicit planning in LLMs and their application to rhyme generation and question answering

Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit

Thought Anchors: Which LLM Reasoning Steps Matter?

Chain-of-Thought Reasoning In The Wild Is Not Always Faithful

Steering Out-of-Distribution Generalization with Concept Ablation Fine-Tuning

Real-Time Detection of Hallucinated Entities in Long-Form Generation

Base Models Know How to Reason, Thinking Models Learn When

Eliciting Secret Knowledge from Language Models

Do I Know This Entity? Knowledge Awareness and Hallucinations in Language Models

Learning Multi-Level Features with Matryoshka Sparse Autoencoders

Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning

Towards Principled Evaluations of Sparse Autoencoders for Interpretability and Control

Sparse Autoencoders Do Not Find Canonical Units of Analysis

Are Sparse Autoencoders Useful? A Case Study in Sparse Probing

Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrieval

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models

SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability

Scaling Sparse Feature Circuits For Studying In-Context Learning

Interpreting Attention Layer Outputs with Sparse Autoencoders

Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders

Scaling Sparse Feature Circuits For Studying In-Context Learning