Research

Fundamental interpretability research to understand and intentionally design advanced AI systems

Fundamental Research

Uncovering Neural Geometry in Vision Models With Block-Sparse Featurizers

Fel et al.
·
July 7, 2026

Uncovering Neural Geometry in Vision Models With Block-Sparse Featurizers

Fel et al.
·
July 7, 2026

Meandering on Manifolds: The Neural Geometry of Stories Over Time

Bigelow et al.
·
June 23, 2026

Meandering on Manifolds: The Neural Geometry of Stories Over Time

Bigelow et al.
·
June 23, 2026

Predictive Data Debugging: Reveal and Shape What Your Model Learns, Before You Train

Bergen et al.
·
June 11, 2026

Predictive Data Debugging: Reveal and Shape What Your Model Learns, Before You Train

Bergen et al.
·
June 11, 2026

Logits as a new monitor for evaluation awareness

Santiago Aranguri
·
June 4, 2026

Logits as a new monitor for evaluation awareness

Santiago Aranguri
·
June 4, 2026

Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention

Huang et al.
·
June 1, 2026

Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention

Huang et al.
·
June 1, 2026

Can SAEs Capture Neural Geometry?

Bhalla et al.
·
May 21, 2026

Can SAEs Capture Neural Geometry?

Bhalla et al.
·
May 21, 2026

A Geometric Calculator Inside a Neural Network

Feucht et al.
·
May 14, 2026

A Geometric Calculator Inside a Neural Network

Feucht et al.
·
May 14, 2026

Predicting Rare LLM Failures with 30× Fewer Rollouts

Aranguri & Pernice
·
May 13, 2026

Predicting Rare LLM Failures with 30× Fewer Rollouts

Aranguri & Pernice
·
May 13, 2026

Steering Along Manifolds to Control Neural Networks

Wurgaft et al.
·
May 7, 2026

Steering Along Manifolds to Control Neural Networks

Wurgaft et al.
·
May 7, 2026

The World Inside Neural Networks

Geiger et al.
·
May 7, 2026

The World Inside Neural Networks

Geiger et al.
·
May 7, 2026

Paper Summary: Interpreting Language Model Parameters

Bushnaq et al.
·
May 5, 2026

Paper Summary: Interpreting Language Model Parameters

Bushnaq et al.
·
May 5, 2026

Interpreting Language Model Parameters

Bushnaq et al.
·
May 5, 2026

Interpreting Language Model Parameters

Bushnaq et al.
·
May 5, 2026

Verbalized Eval Awareness Inflates Measured Safety

Aranguri and Bloom
·
May 4, 2026

Verbalized Eval Awareness Inflates Measured Safety

Aranguri and Bloom
·
May 4, 2026

Probe-Based Data Attribution: Surfacing and Mitigating Undesirable Behaviors in LLM Post-Training

Xiao and Aranguri
·
April 29, 2026

Probe-Based Data Attribution: Surfacing and Mitigating Undesirable Behaviors in LLM Post-Training

Xiao and Aranguri
·
April 29, 2026

Explaining 4.2 million genetic variants with state-of-the-art, interpretable predictions

Pearce et al.
·
April 14, 2026

Explaining 4.2 million genetic variants with state-of-the-art, interpretable predictions

Pearce et al.
·
April 14, 2026

Covariance-based Sequence Pooling

Dooms et al.
·
April 10, 2026

Covariance-based Sequence Pooling

Dooms et al.
·
April 10, 2026

Using Self-Correcting Search to Accelerate Materials Discovery

Hazra et al.
·
April 1, 2026

Using Self-Correcting Search to Accelerate Materials Discovery

Hazra et al.
·
April 1, 2026

Reasoning Theater: Probing for Performative Chain-of-Thought

Boppana et al.
·
March 12, 2026

Reasoning Theater: Probing for Performative Chain-of-Thought

Boppana et al.
·
March 12, 2026

Features as Rewards: Using Interpretability to Reduce Hallucinations

Prasad et al.
·
February 11, 2026

Features as Rewards: Using Interpretability to Reduce Hallucinations

Prasad et al.
·
February 11, 2026

Using Interpretability to Identify a Novel Class of Alzheimer's Biomarkers

Wang et al.
·
January 28, 2026

Using Interpretability to Identify a Novel Class of Alzheimer's Biomarkers

Wang et al.
·
January 28, 2026

Understanding Memorization via Loss Curvature

Merullo et al.
·
November 6, 2025

Understanding Memorization via Loss Curvature

Merullo et al.
·
November 6, 2025

Priors in Time: Missing Inductive Biases for Language Model Interpretability

Lubana et al.
·
November 3, 2025

Priors in Time: Missing Inductive Biases for Language Model Interpretability

Lubana et al.
·
November 3, 2025

Belief Dynamics Reveal the Dual Nature of In-Context Learning and Activation Steering

Bigelow et al.
·
November 1, 2025

Belief Dynamics Reveal the Dual Nature of In-Context Learning and Activation Steering

Bigelow et al.
·
November 1, 2025

Deploying Interpretability to Production with Rakuten: SAE Probes for PII Detection

Nguyen et al.
·
October 28, 2025

Deploying Interpretability to Production with Rakuten: SAE Probes for PII Detection

Nguyen et al.
·
October 28, 2025

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

Gur-Arieh et al.
·
October 7, 2025

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

Gur-Arieh et al.
·
October 7, 2025

Understanding Sparse Autoencoder Scaling in the Presence of Feature Manifolds

Michaud et al.
·
September 4, 2025

Understanding Sparse Autoencoder Scaling in the Presence of Feature Manifolds

Michaud et al.
·
September 4, 2025

Finding the Tree of Life in Evo 2

Pearce et al.
·
August 28, 2025

Finding the Tree of Life in Evo 2

Pearce et al.
·
August 28, 2025

Adversarial Examples Are Not Bugs, They Are Superposition

Gorton & Lewis
·
August 26, 2025

Adversarial Examples Are Not Bugs, They Are Superposition

Gorton & Lewis
·
August 26, 2025

Discovering Undesired Rare Behaviors via Model Diff Amplification

Aranguri & McGrath
·
August 21, 2025

Discovering Undesired Rare Behaviors via Model Diff Amplification

Aranguri & McGrath
·
August 21, 2025

The Circuits Research Landscape: Results and Perspectives

Lindsey et al.
·
August 5, 2025

The Circuits Research Landscape: Results and Perspectives

Lindsey et al.
·
August 5, 2025

Towards Scalable Parameter Decomposition

Bushnaq et al.
·
June 28, 2025

Towards Scalable Parameter Decomposition

Bushnaq et al.
·
June 28, 2025

Replicating Circuit Tracing for a Simple Known Mechanism

Loeffler et al.
·
June 11, 2025

Replicating Circuit Tracing for a Simple Known Mechanism

Loeffler et al.
·
June 11, 2025

Painting With Concepts Using Diffusion Model Latents

Cammarata et al.
·
May 27, 2025

Painting With Concepts Using Diffusion Model Latents

Cammarata et al.
·
May 27, 2025

Under the Hood of a Reasoning Model

Hazra et al.
·
April 15, 2025

Under the Hood of a Reasoning Model

Hazra et al.
·
April 15, 2025

Interpreting Evo 2: Arc Institute's Next-Generation Genomic Foundation Model

Gorton et al.
·
February 20, 2025

Interpreting Evo 2: Arc Institute's Next-Generation Genomic Foundation Model

Gorton et al.
·
February 20, 2025

Open Problems in Mechanistic Interpretability

Sharkey et al.
·
January 27, 2025

Open Problems in Mechanistic Interpretability

Sharkey et al.
·
January 27, 2025

Mapping the Latent Space of Llama 3.3 70B

McGrath et al.
·
December 23, 2024

Mapping the Latent Space of Llama 3.3 70B

McGrath et al.
·
December 23, 2024

Understanding and Steering Llama 3 with Sparse Autoencoders

McGrath et al.
·
September 25, 2024

Understanding and Steering Llama 3 with Sparse Autoencoders

McGrath et al.
·
September 25, 2024