Research

.

Towards an Understanding of Intelligence

Fingerprinting All AI Cluster I/O Without Mutually Trusted Processors

Naci Cankaya, Jakub Krys, Jonathan Ng, Luke Marks, Felix Kruckel

Ablation-Reversible Heads Don't Transfer: A Stress Test for Mechanistic Role Claims in Transformers

STRIDE: Training Data Attribution via Sparse Recovery from Subset Perturbations

Rishit Dagli, Abir Harrasse, Luke Zhang, Florent Draye, Amirali Abdullah, Bernhard Scholkopf, Zhijing Jin

Riemannian-Manifold Steering: Geometry-Aware Generative Autoencoders for Representation Control

Narmeen Oozeer, Shivam Raval, Philip Quirke, Manikandan Ravikiran, Jeff Phillips, Shriyash Upadhyay, Amirali Abdullah

Make Mechanistic Interpretability Auditable: A Call to Develop Guidelines via Continuous Collaborative Reviewing

Michael Lan, Narmeen Fatimah Oozeer, Chaithanya Bandi, Philip Quirke, Austin Meek, Fazl Barez, Amirali Abdullah

Geometry-Aware CLIP Retrieval via Local Cross-Modal Alignment and Steering

Nirmalendu Prakash, Narmeen Fatimah Oozeer, Xin Su, Phillip Howard, Shaan Shah, Zoe Wanying He, Shuang Wu, Shivam Raval, Roy Ka-Wei Lee, Meenakshi Khosla, Amir Abdullah

Curveball Steering: The Right Direction To Steer Isn't Always Linear

Shivam Raval, Hae Jin Song, Linlin Wu, Abir Harrasse, Jeff M. Phillips, Fazl Barez, Amirali Abdullah

Understanding and Mitigating Dataset Corruption in LLM Steering

Cullen Anderson, Narmeen Oozeer, Foad Namjoo, Remy Ogasawara, Amirali Abdullah, Jeff M. Phillips

DreamReader: An Interpretability Toolkit for Text-to-Image Models

Nirmalendu Prakash, Narmeen Oozeer, Michael Lan, Luka Samkharadze, Phillip Howard, Roy Ka-Wei Lee, Dhruv Nathawani, Shivam Raval, Amirali Abdullah

Old Habits Die Hard: How Conversational History Geometrically Traps LLMs

Adi Simhi, Fazl Barez, Martin Tutek, Yonatan Belinkov, Shay B. Cohen

Spectral Superposition: A Theory of Feature Geometry

Georgi Ivanov, Narmeen Oozeer, Shivam Raval, Tasana Pejovic, Shriyash Upadhyay, Amir Abdullah

Are LLM Evaluators Really Narcissists? Sanity Checking Self-Preference in LLMs

Dani Roytburg, Matthew Bozoukov, Matthew Nguyen, Jou Barzdukas, Mackenzie Puig-Hall, Narmeen Oozeer

Position: Require Frontier AI Labs To Release Small 'Analog' Models

Shriyash Upadhyay, Chaithanya Bandi, Narmeen Oozeer, Philip Quirke

Beyond Linear Probes: Dynamic Safety Monitoring for Language Models

James Oldfield, Philip Torr, Ioannis Patras, Adel Bibi, Fazl Barez

Query Circuits: Explaining How Language Models Answer User Prompts

Tung-Yu Wu, Fazl Barez

Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer

Simon Schrodi, Elias Kempf, Fazl Barez, Thomas Brox

Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLMs

Dani Roytburg, Matthew Bozoukov, Matthew Nguyen, Jou Barzdukas, Simon Fu, Narmeen Oozeer

Distribution-Aware Feature Selection for SAEs

Narmeen Oozeer, Nirmalendu Prakash, Michael Lan, Alice Rigg, Amirali Abdullah

Method for Converting Models to Programs

Shriyash K. Upadhyay, Etan J. Ginsberg, Luka Samkharadze

Model Mapping: Can We Turn Transformers into Programs?

Shriyash Upadhyay, Benjamin Keigwin, Chaithanya Bandi, Yiwei Wu, Luka Samkharadze, Jason Hu, Etan Ginsberg

Beyond Linear Steering: Unified Multi-Attribute Control for Language Models

Narmeen Oozeer, Luke Marks, Shreyans Jain, Fazl Barez, Amirali Abdullah

Beyond Monoliths: Expert Orchestration for More Capable, Democratic, and Safe LLMs

Philip Quirke, Narmeen Oozeer, Chaithanya Bandi, Amir Abdullah, Jason Hoelscher-Obermaier, Jeff M. Phillips, Joshua Greaves, Clement Neo, Michael Lan, Fazl Barez, Shriyash Upadhyay

TinySQL: A Progressive Text-to-SQL Dataset for Mechanistic Interpretability

Abir Harrasse, Philip Quirke, Clement Neo, Dhruv Nathawani, Luke Marks, Amir Abdullah

Activation Space Interventions Can Be Transferred Between Large Language Models

Narmeen Oozeer, Dhruv Nathawani, Nirmalendu Prakash, Michael Lan, Abir Harrasse, Amirali Abdullah

Bilinear Convolution Decomposition for Causal RL Interpretability

Narmeen Oozeer, Sinem Erisken, Alice Rigg

Debate, Deliberate, Decide (D3): A Cost-Aware Adversarial Framework for Reliable and Interpretable LLM Evaluation

Abir Harrasse, Chaithanya Bandi, Hari Bandi

Understanding Addition and Subtraction in Transformers

Philip Quirke, Clement Neo, Fazl Barez

The Capability Frontier: Benchmarks Miss 82% of Model Performance

Narmeen Oozeer, Philip Quirke, Fazl Barez