Token Taxes Can Mitigate AI's Economic Risks

L. Irwin, T.-Y. Wu, F. Barez

Old Habits Die Hard: How Conversational History Geometrically Traps LLMs

A. Simhi, F. Barez, M. Tutek, Y. Belinkov, S. B. Cohen

AutoControl Arena: Synthesizing Executable Test Environments for Frontier AI Risk Evaluation

C. Li, P. Lu, X. Pan, F. Barez, M. Yang

Same Answer, Different Representations: Hidden Instability in VLMs

F. A. Wani, A. Suglia, R. Saxena, A. P. Gema, W. C. Kwan, F. Barez, et al.

Interpretability Can Be Actionable

H. Orgad, F. Barez, T. Haklay, I. Lee, M. Mosbach, A. Reusch, N. Saphra, et al.

When AI Systems Learn During Deployment, Our Safety Evaluations Break

F. Barez

What AI Governance Needs from Mechanistic Auditing

Uplifting Human Decision Making in AI Evaluation by Automating Benchmark Design

There Are Futures That Benchmark-Driven AI Cannot See

S. Lotfi, A. Iranmanesh, L. Naghashyar, A. Shirali, F.N. Haredasht, S. Koyejo, P. Torr, Y.S. Lee, F. Barez, J. Lehman, P. Norvig, A. Narayanan

The Capability Frontier: Benchmarks Miss 82% of Model Performance

B. Fowler, R. Smith, D. T. Graviet, W. Myers, J. Greaves, N. F. Oozeer, A. García, et al.

Rigorous Interpretation Is a Form of Evaluation

Quantifying the Effect of Test Set Contamination on Generative Evaluations

R. Schaeffer, J. Kazdan, B. Abbasi, K. Z. Liu, B. Miranda, A. Ahmed, F. Barez, et al.

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

M. Lan, N.F. Oozeer, C. Bandi, P. Quirke, A. Meek, F. Barez, A. Abdullah

Don't Just "Fix it in Post": A Science of AI Must Study Learning Dynamics

S. Biderman, M.A. Khan, N. Mireshghallah, C. Arnett, F. Barez, N. Saphra

Beyond Alignment: Why Robotic Foundation Models Need Context-Aware Safety

A. Robey, Z. Ravichandran, E.K. Jones, J. Perlo, F. Barez, V. Kumar, J.Z. Kolter, H. Hassani, G.J. Pappas

Automated Interpretability-Driven Model Auditing and Control: A Research Agenda

F. Barez

Agentic Product Maturity Ladder V0.1

S. McGregor, D. Nathani, L. Saouma, F. Barez, A. Foundjem, et al.

Full-Stack Alignment: Co-Aligning AI and Institutions with Thicker Models of Value

R. Lowe, J. Edelman, T. Zhi-Xuan, O. Klingefjord, E. Hain, V. Wang, A. Sarkar, F. Barez, et al.

Establishing Best Practices for Building Rigorous Agentic Benchmarks

Y. Zhu, T. Jin, Y. Pruksachatkun, A. Zhang, S. Liu, S. Cui, S. Kapoor, F. Barez, et al.

Emerging Risks from Embodied AI Require Urgent Policy Action

J. Perlo, A. Robey, F. Barez, J. Mökander

Context Matters: Analyzing the Generalizability of Linear Probing and Steering Across Diverse Scenarios

I. Agarwal, S. Navani, F. Barez

Trust Me, I'm Wrong: High-Certainty Hallucinations in LLMs

A. Simhi, I. Itzhak, F. Barez, G. Stanovsky, Y. Belinkov

Same Question, Different Words: A Latent Adversarial Framework for Prompt Robustness

T. Fu, F. Barez

Precise In-Parameter Concept Erasure in Large Language Models

Y. Gur-Arieh, C. Suslik, Y. Hong, F. Barez, M. Geva

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

N. Oozeer, L. Marks, F. Barez, A. Abdullah

Val-Bench: Measuring Value Alignment in Language Models

A. Gupta, D. O'Shea, F. Barez

Rethinking Safety in LLM Fine-Tuning: An Optimization Perspective

M. Kim, J. M. Kwak, L. Alssum, B. Ghanem, P. Torr, D. Krueger, F. Barez†, A. Bibi†

HACK: Hallucinations Along Certainty and Knowledge Axes

A. Simhi, J. Herzig, I. Itzhak, D. Arad, Z. Gekhman, R. Reichart, F. Barez, et al.

Chain-of-Thought Hijacking

J. Zhao, T. Fu, R. Schaeffer, M. Sharma, F. Barez

Towards Understanding Subliminal Learning: When and How Hidden Biases Transfer

S. Schrodi, E. Kempf, F. Barez, T. Brox

Query Circuits: Explaining How Language Models Answer User Prompts

T.-Y. Wu, F. Barez

Beyond Linear Probes: Dynamic Safety Monitoring for Language Models

J. Oldfield, P. Torr, I. Patras, A. Bibi, F. Barez

Scaling Sparse Feature Circuit Finding for In-Context Learning

D. Kharlapenko, S. Shabalin, F. Barez, A. Conmy, N. Nanda

PoisonBench: Assessing Large Language Model Vulnerability to Data Poisoning

T. Fu*, M. Sharma, P. Torr, S. B. Cohen, D. Krueger, F. Barez*

Do Sparse Autoencoders Generalize? A Case Study of Answerability

L. Heindrich, P. Torr, F. Barez, V. Thost

The Singapore Consensus on Global AI Safety Research Priorities

Y. Bengio, T. Maharaj, L. Ong, S. Russell, D. Song, M. Tegmark, L. Xue, F. Barez, et al.

In Which Areas of Technical AI Safety Could Geopolitical Rivals Cooperate?

B. Bucknall, S. Siddiqui, L. Thurnherr, C. McGurk, B. Harack, A. Reuel, F. Barez, et al.

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

P. Quirke, N. Oozeer, C. Bandi, A. Abdullah, J. Hoelscher-Obermaier, F. Barez, et al.

SafetyNet: Detecting Harmful Outputs in LLMs by Modeling and Monitoring Deceptive Behaviors

M. Chaudhary, F. Barez

Towards Interpreting Visual Information Processing in Vision-Language Models

C. Neo, L. Ong, P. Torr, M. Geva, D. Krueger, F. Barez

Rethinking AI Cultural Alignment

M. Bravansky, F. Trhlík, F. Barez

AILuminate: Introducing v1.0 of the AI Risk and Reliability Benchmark from MLCommons

S. Ghosh, H. Frase, A. Williams, S. Luger, P. Röttger, F. Barez, S. McGregor, et al.

Chain-of-Thought Is Not Explainability

F. Barez, T.-Y. Wu, I. Arcuschin, M. Lan, V. Wang, N. Siegel, N. Collignon, C. Neo, I. Lee, A. Paren, A. Bibi, R. Trager, D. Fornasiere, J. Yan, Y. Elazar, Y. Bengio

Verification for International AI Governance

B. Harack, R. Trager, A. Reuel, D. Manheim, M. Brundage, O. Aarne, et al.

Toward Resisting AI-Enabled Authoritarianism

F. Barez, I. Friend, K. Reid, I. Krawczuk, V. Wang, J. Mökander, P. Torr, J. Morse, R. Trager

Safety Frameworks and Standards: A Comparative Analysis to Advance Risk Management of Frontier AI

M. Ziosi, J. Gealy, M. Plueckebaum, D. Kossack, S. Campos, L. Saouma, F. Barez, et al.

Plan B: Training LLMs to Fail Less Severely

J. Stastny, N. Warncke, D. Xu, A. Lynch, F. Barez, H. Sleight, E. Perez

Open Problems in Machine Unlearning for AI Safety

F. Barez, T. Fu, A. Prabhu, S. Casper, A. Sanyal, A. Bibi, A. O'Gara, R. Kirk, B. Bucknall, T. Fist, L. Ong, P. Torr, K.-Y. Lam, R. Trager, D. Krueger, S. Mindermann, J. Hernández-Orallo, M. Geva, Y. Gal

Jailbreak Defense in a Narrow Domain: Limitations of Existing Methods and a New Transcript-Classifier Approach

T. T. Wang, J. Hughes, H. Sleight, R. Schaeffer, R. Agrawal, F. Barez, et al.

Interpreting Learned Feedback Patterns in Large Language Models

L. Marks*, A. Abdullah*, C. Neo, R. Arike, D. Krueger, P. Torr, F. Barez*

Best-of-N Jailbreaking

J. Hughes, S. Price, A. Lynch, R. Schaeffer, F. Barez, S. Koyejo, H. Sleight, E. Jones, E. Perez

Towards Interpretable Sequence Continuation: Analyzing Shared Circuits in Large Language Models

M. Lan, P. Torr, F. Barez

Interpreting Context Look-ups in Transformers: Investigating Attention-MLP Interactions

C. Neo*, S. B. Cohen, F. Barez*

Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoders

L. Marks, A. Paren, D. Krueger, F. Barez

Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders

M. Lan, P. Torr, A. Meek, D. Krueger, F. Barez

Large Language Models Relearn Removed Concepts

M. Lo*, S. B. Cohen, F. Barez*

Visualizing Neural Network Imagination

N. Wichers, V. Tao, R. Volpato, F. Barez

The Scaling Behavior of Large Language Models

A. V. Miceli-Barone, F. Barez, S. B. Cohen, E. Voita, U. Germann, M. Lukasik

Position: Near to Mid-Term Risks and Opportunities of Open-Source Generative AI

F. Eiras, A. Petrov, B. Vidgen, C. S. de Witt, F. Pizzati, K. Elkins, F. Barez, et al.

Mechanistic Interpretability Workshop at ICML 2024

F. Barez, M. Geva, L. Chan, A. Geiger, K. Yin, N. Nanda, et al.

Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models

C. Denison, M. MacDiarmid, F. Barez, D. Duvenaud, S. Kravec, S. Marks, et al.

Understanding Addition in Transformers

P. Quirke, F. Barez

Increasing Trust in Language Models Through the Reuse of Verified Circuits

P. Quirke, C. Neo, F. Barez

What Does GPT Store in Its MLP Weights? A Case Study of Long-Range Dependencies

T. Clark, S. B. Cohen, F. Barez

Sleeper Agents: Training Deceptive LLMs That Persist Through Safety Training

E. Hubinger, C. Denison, J. Mu, M. Lambert, M. Tong, M. MacDiarmid, F. Barez, et al.

Safeguarding AI in Finance: Lessons for Regulated Industries

F. Barez, L. Marks

Measuring Value Alignment

F. Barez, P. Torr

DeepDecipher: Accessing and Investigating Neuron Activation in Large Language Models

A. Garde, E. Kran, F. Barez

AI Systems of Concern

K. Matteucci, S. Avin, F. Barez, S. Ó hÉigeartaigh

The Larger They Are, the Harder They Fail: Language Models Do Not Recognize Identifier Swaps in Python

A. V. M. Barone*, F. Barez*, I. Konstas, S. B. Cohen

Detecting Edit Failures in Large Language Models: An Improved Specificity Benchmark

J. Hoelscher-Obermaier*, J. Persson*, E. Kran, I. Konstas, F. Barez*

Neuron to Graph: Interpreting Language Model Neurons at Scale

A. Foote*, N. Nanda, E. Kran, I. Konstas, S. Cohen, F. Barez*

Value-Evolutionary-Based Reinforcement Learning

P. Li, J. Hao, H. Tang, Y. Zheng, F. Barez

The Alan Turing Institute's Response to the House of Lords Large Language Models Call for Evidence

F. Barez, P. H. S. Torr, A. Petrov, C. Ashurst, J. Ding, A. Janjeva, A. Babuta, et al.

Identifying a Preliminary Circuit for Predicting Gendered Pronouns in GPT-2 Small

C. Mathwin, G. Corlouer, E. Kran, F. Barez, N. Nanda

Fairness in AI and Its Long-Term Implications on Society

O. Bohdal*, T. Hospedales, P. H. S. Torr, F. Barez*

Exploring the Advantages of Transformers for High-Frequency Trading

F. Barez, P. Bilokon, A. Gervais, N. Lisitsyn

Data Study Group Final Report: Keep Wales Tidy and Keep Scotland Beautiful

F. Barez, H. Ni, S. Fielding, A. Rehman, O. Babajide, M. H. Modi, M. O. Akinsolu

System III: Learning with Domain Knowledge for Safety Constraints

F. Barez, H. Hasanbieg, A. Abbate

PMIC: Improving Multi-Agent RL with Progressive Mutual Information Collaboration

P. Li, H. Tang, T. Yang, X. Hao, T. Sang, Y. Zheng, J. Hao, M. E. Taylor, Z. Wang, F. Barez

ED2: An Environment Dynamics Decomposition Framework

C. Wang, T. Yang, Y. Zheng, H. Tang, F. Barez, J. Liu, J. Peng, Z. Sun

Discovering Topics and Trends in the UK Government Web Archive

D. Beavan, F. Barez, M. Bel, J. Fitzgerald, E. Goudarouli, K. Kollnig, et al