Cover Image

Open Problems in Machine Unlearning for AI Safety

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Barez, Fazl, Fu, Tingchen, Prabhu, Ameya, Casper, Stephen, Sanyal, Amartya, Bibi, Adel, O'Gara, Aidan, Kirk, Robert, Bucknall, Ben, Fist, Tim, Ong, Luke, Torr, Philip, Lam, Kwok-Yan, Trager, Robert, Krueger, David, Mindermann, Sören, Hernandez-Orallo, José, Geva, Mor, Gal, Yarin
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence Computers and Society
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Towards Interpreting Visual Information Processing in Vision-Language Models
by: Neo, Clement, et al.
Published: (2024)

Same Question, Different Words: A Latent Adversarial Framework for Prompt Robustness
by: Fu, Tingchen, et al.
Published: (2025)

PoisonBench: Assessing Large Language Model Vulnerability to Data Poisoning
by: Fu, Tingchen, et al.
Published: (2024)

Beyond Linear Probes: Dynamic Safety Monitoring for Language Models
by: Oldfield, James, et al.
Published: (2025)

Corrective Machine Unlearning
by: Goel, Shashwat, et al.
Published: (2024)

Rethinking Safety in LLM Fine-tuning: An Optimization Perspective
by: Kim, Minseon, et al.
Published: (2025)

Precise In-Parameter Concept Erasure in Large Language Models
by: Gur-Arieh, Yoav, et al.
Published: (2025)

Towards Interpretable Sequence Continuation: Analyzing Shared Circuits in Large Language Models
by: Lan, Michael, et al.
Published: (2023)

Chain-of-Thought Hijacking
by: Zhao, Jianli, et al.
Published: (2025)

Efficient Lifelong Model Evaluation in an Era of Rapid Progress
by: Prabhu, Ameya, et al.
Published: (2024)

Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders
by: Lan, Michael, et al.
Published: (2024)

Do Sparse Autoencoders Generalize? A Case Study of Answerability
by: Heindrich, Lovis, et al.
Published: (2025)

Enhancing Neural Network Interpretability with Feature-Aligned Sparse Autoencoders
by: Marks, Luke, et al.
Published: (2024)

Interpreting Learned Feedback Patterns in Large Language Models
by: Marks, Luke, et al.
Published: (2023)

SafetyNet: Detecting Harmful Outputs in LLMs by Modeling and Monitoring Deceptive Behaviors
by: Chaudhary, Maheep, et al.
Published: (2025)

Understanding Addition in Transformers
by: Quirke, Philip, et al.
Published: (2023)

Detecting LLM Hallucination Through Layer-wise Information Deficiency: Analysis of Ambiguous Prompts and Unanswerable Questions
by: Kim, Hazel, et al.
Published: (2024)

Hemicrania continua exacerbations with prodromal burping as a potential autonomic symptom
by: Raluca A. Negulescu, et al.
Published: (2026)

Position: Ensuring mutual privacy is necessary for effective external evaluation of proprietary AI systems
by: Bucknall, Ben, et al.
Published: (2025)

MIP against Agent: Malicious Image Patches Hijacking Multimodal OS Agents
by: Aichberger, Lukas, et al.
Published: (2025)

Query Circuits: Explaining How Language Models Answer User Prompts
by: Wu, Tung-Yu, et al.
Published: (2025)

Actor–Partner Model of Parenting and Co‐Parenting Practices and Youth Resilience During the COVID‐19 Pandemic
by: Olivia Aspiras, et al.
Published: (2025)

No "Zero-Shot" Without Exponential Data: Pretraining Concept Frequency Determines Multimodal Model Performance
by: Udandarao, Vishaal, et al.
Published: (2024)

From Categories to Classifiers: Name-Only Continual Learning by Exploring the Web
by: Prabhu, Ameya, et al.
Published: (2023)

Biased rate estimates in bump-hunt searches
by: Murray, William, et al.
Published: (2025)

Rethinking AI Cultural Alignment
by: Bravansky, Michal, et al.
Published: (2025)

Understanding Addition and Subtraction in Transformers
by: Quirke, Philip, et al.
Published: (2024)

OMNI-LEAK: Orchestrator Multi-Agent Network Induced Data Leakage
by: Naik, Akshat, et al.
Published: (2026)

Delta-Influence: Unlearning Poisons via Influence Functions
by: Li, Wenjie, et al.
Published: (2024)

Prompting a Pretrained Transformer Can Be a Universal Approximator
by: Petrov, Aleksandar, et al.
Published: (2024)

When Do Prompting and Prefix-Tuning Work? A Theory of Capabilities and Limitations
by: Petrov, Aleksandar, et al.
Published: (2023)

Activities of Garlic Oil, Garlic Powder, and Their Diallyl Constituents against Helicobacter pylori
by: O'Gara E.A. Hilld.J and Maslin D.J
Published: (2000)

Token Taxes: mitigating AGI's economic risks
by: Irwin, Lucas, et al.
Published: (2026)

Large Language Models Relearn Removed Concepts
by: Lo, Michelle, et al.
Published: (2024)

VAL-Bench: Belief Consistency as a measure for Value Alignment in Language Models
by: Gupta, Aman, et al.
Published: (2025)

Interpreting Context Look-ups in Transformers: Investigating Attention-MLP Interactions
by: Neo, Clement, et al.
Published: (2024)

Protecting against simultaneous data poisoning attacks
by: Alex, Neel, et al.
Published: (2024)

Estimating Knowledge in Large Language Models Without Generating a Single Token
by: Gottesman, Daniela, et al.
Published: (2024)

Inferring Functionality of Attention Heads from their Parameters
by: Elhelo, Amit, et al.
Published: (2024)

In Which Areas of Technical AI Safety Could Geopolitical Rivals Cooperate?
by: Bucknall, Ben, et al.
Published: (2025)