Cover Image

Deep Ignorance: Filtering Pretraining Data Builds Tamper-Resistant Safeguards into Open-Weight LLMs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	O'Brien, Kyle, Casper, Stephen, Anthony, Quentin, Korbak, Tomek, Kirk, Robert, Davies, Xander, Mishra, Ishan, Irving, Geoffrey, Gal, Yarin, Biderman, Stella
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Fundamental Limitations in Pointwise Defences of LLM Finetuning APIs
by: Davies, Xander, et al.
Published: (2025)

How to evaluate control measures for LLM agents? A trajectory from today to superintelligence
by: Korbak, Tomek, et al.
Published: (2025)

Boundary Point Jailbreaking of Black-Box LLMs
by: Davies, Xander, et al.
Published: (2026)

Safety Cases: A Scalable Approach to Frontier AI Safety
by: Hilton, Benjamin, et al.
Published: (2025)

A sketch of an AI control safety case
by: Korbak, Tomek, et al.
Published: (2025)

Tamper-Resistant Safeguards for Open-Weight LLMs
by: Tamirisa, Rishub, et al.
Published: (2024)

Practical challenges of control monitoring in frontier AI deployments
by: Lindner, David, et al.
Published: (2025)

STACK: Adversarial Attacks on LLM Safeguard Pipelines
by: McKenzie, Ian R., et al.
Published: (2025)

An Example Safety Case for Safeguards Against Misuse
by: Clymer, Joshua, et al.
Published: (2025)

Safety case template for frontier AI: A cyber inability argument
by: Goemans, Arthur, et al.
Published: (2024)

Lessons from Studying Two-Hop Latent Reasoning
by: Balesni, Mikita, et al.
Published: (2024)

Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities
by: Che, Zora, et al.
Published: (2025)

Existing Large Language Model Unlearning Evaluations Are Inconclusive
by: Feng, Zhili, et al.
Published: (2025)

Training Agents to Self-Report Misbehavior
by: Lee, Bruce W., et al.
Published: (2026)

Safeguarding Your Business and Management Succession.
by: Maness, Irving, et al.
Published: (1965)

Expanding External Access To Frontier AI Models For Dangerous Capability Evaluations
by: Charnock, Jacob, et al.
Published: (2026)

RedCodeAgent: Automatic Red-teaming Agent against Diverse Code Agents
by: Guo, Chengquan, et al.
Published: (2025)

MITS-GAN: Safeguarding Medical Imaging from Tampering with Generative Adversarial Networks
by: Pasqualino, Giovanni, et al.
Published: (2024)

Selective Safety Steering via Value-Filtered Decoding
by: Einbinder, Bat-Sheva, et al.
Published: (2026)

Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples
by: Souly, Alexandra, et al.
Published: (2025)

TamperBench: Systematically Stress-Testing LLM Safety Under Fine-Tuning and Tampering
by: Hossain, Saad, et al.
Published: (2026)

SeCodePLT: A Unified Platform for Evaluating the Security of Code GenAI
by: Nie, Yuzhou, et al.
Published: (2024)

Alignment Pretraining: AI Discourse Causes Self-Fulfilling (Mis)alignment
by: Tice, Cameron, et al.
Published: (2026)

Language Models Change Facts Based on the Way You Talk
by: Kearney, Matthew, et al.
Published: (2025)

Do Multilingual LLMs Think In English?
by: Schut, Lisa, et al.
Published: (2025)

In-Context Learning Learns Label Relationships but Is Not Conventional Learning
by: Kossen, Jannik, et al.
Published: (2023)

Simple Baselines are Competitive with Code Evolution
by: Gideoni, Yonatan, et al.
Published: (2026)

The Benefits and Risks of Transductive Approaches for AI Fairness
by: Razzak, Muhammed, et al.
Published: (2024)

Safeguarding Data Integrity in Online Podiatry Research: Understanding and Managing Non‐Genuine Participation
by: Will Gurr, et al.
Published: (2026)

Aligning language models with human preferences
by: Korbak, Tomasz
Published: (2024)

Weighted Conditional Flow Matching
by: Calvo-Ordonez, Sergio, et al.
Published: (2025)

Evaluating whether AI models would sabotage AI safety research
by: Kirk, Robert, et al.
Published: (2026)

UK AISI Alignment Evaluation Case-Study
by: Souly, Alexandra, et al.
Published: (2026)

Adaptable Intelligent Filters of Dual Thermo‐ and pH‐ Responsive Filter Material
by: Jingwei Wu, et al.
Published: (2024)

Red novae, stellar mergers in binary and triple systems, and bipolar nebulae
by: Kaminski, Tomek
Published: (2024)

Emergent Compositional Communication for Latent World Properties
by: Kaszyński, Tomek
Published: (2026)

In Defiance: 20 Abolitionists You Were Never Taught in Schools. By TomWeiner and AmilcarShabazz. Northampton, MA: Interlink/Olive Branch Press, 2025. 248 pp. $25.00 (paperback). ISBN: 978‐1‐62‐371661‐5
by: Beverly Tomek
Published: (2025)

Stabilizing Policy Gradients for Sample-Efficient Reinforcement Learning in LLM Reasoning
by: Melo, Luckeciano C., et al.
Published: (2025)

TextCAVs: Debugging vision models using text
by: Nicolson, Angus, et al.
Published: (2024)

Temporal-Difference Variational Continual Learning
by: Melo, Luckeciano C., et al.
Published: (2024)