Cover Image

Beyond Data Filtering: Knowledge Localization for Capability Removal in LLMs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Shilov, Igor, Cloud, Alex, Gema, Aryo Pradipta, Goldman-Wetzler, Jacob, Panickssery, Nina, Sleight, Henry, Jones, Erik, Anil, Cem
Format:	Preprint
Published:	2025
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

The Hot Mess of AI: How Does Misalignment Scale With Model Intelligence and Task Complexity?
by: Hägele, Alexander, et al.
Published: (2026)

Gradient Routing: Masking Gradients to Localize Computation in Neural Networks
by: Cloud, Alex, et al.
Published: (2024)

Lost in Time: Clock and Calendar Understanding Challenges in Multimodal LLMs
by: Saxena, Rohit, et al.
Published: (2025)

Inverse Scaling in Test-Time Compute
by: Gema, Aryo Pradipta, et al.
Published: (2025)

Scalpel vs. Hammer: GRPO Amplifies Existing Capabilities, SFT Replaces Them
by: Rajani, Neel, et al.
Published: (2025)

Edinburgh Clinical NLP at SemEval-2024 Task 2: Fine-tune your model unless you have access to GPT-4
by: Gema, Aryo Pradipta, et al.
Published: (2024)

Parameter-Efficient Fine-Tuning of LLaMA for the Clinical Domain
by: Gema, Aryo Pradipta, et al.
Published: (2023)

CoMAT: Chain of Mathematically Annotated Thought Improves Mathematical Reasoning
by: Leang, Joshua Ong Jun, et al.
Published: (2024)

Edinburgh Clinical NLP at MEDIQA-CORR 2024: Guiding Large Language Models with Hints
by: Gema, Aryo Pradipta, et al.
Published: (2024)

Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineering
by: Zhao, Yu, et al.
Published: (2024)

Distillation Robustifies Unlearning
by: Lee, Bruce W., et al.
Published: (2025)

SCOPE: Self-Play via Co-Evolving Policies for Open-Ended Tasks
by: Kwan, Wai-Chung, et al.
Published: (2026)

Inspection and Control of Self-Generated-Text Recognition Ability in Llama3-8b-Instruct
by: Ackerman, Christopher, et al.
Published: (2024)

Speeding up and reducing memory usage for scientific machine learning via mixed precision
by: Hayford, Joel, et al.
Published: (2024)

Mitigating Many-Shot Jailbreaking
by: Ackerman, Christopher M., et al.
Published: (2025)

Self-Training Large Language Models for Tool-Use Without Demonstrations
by: Luo, Ne, et al.
Published: (2025)

DeCoRe: Decoding by Contrasting Retrieval Heads to Mitigate Hallucinations
by: Gema, Aryo Pradipta, et al.
Published: (2024)

Analysing the Residual Stream of Language Models Under Knowledge Conflicts
by: Zhao, Yu, et al.
Published: (2024)

GRADA: Graph-based Reranking against Adversarial Documents Attack
by: Zheng, Jingjie, et al.
Published: (2025)

Noiser: Bounded Input Perturbations for Attributing Large Language Models
by: Madani, Mohammad Reza Ghasemi, et al.
Published: (2025)

Understanding Jailbreak Success: A Study of Latent Space Dynamics in Large Language Models
by: Ball, Sarah, et al.
Published: (2024)

Can GPT-3.5 Generate and Code Discharge Summaries?
by: Falis, Matúš, et al.
Published: (2024)

A Comparative Study on Patient Language across Therapeutic Domains for Effective Patient Voice Classification in Online Health Discussions
by: Lysandrou, Giorgos, et al.
Published: (2024)

An Analysis of Decoding Methods for LLM-based Agents for Faithful Multi-Hop Question Answering
by: Murphy, Alexander, et al.
Published: (2025)

Unsupervised Elicitation of Language Models
by: Wen, Jiaxin, et al.
Published: (2025)

Analyzing Probabilistic Methods for Evaluating Agent Capabilities
by: Højmark, Axel, et al.
Published: (2024)

Abstractive Red-Teaming of Language Model Character
by: Rahn, Nate, et al.
Published: (2026)

Feminist mental health activism in England c. 1968–95. By K.Mahoney, Manchester: Manchester University Press. 2023. pp. 280. £85.00 (cloth), £85.00 (ebk). ISBN: 9781526162267
by: Sara Wetzler
Published: (2024)

Fight Books in Comparative Perspective. An Introduction
by: Sixt Wetzler
Published: (2020)

Paul Bowman, Mythologies of Martial Arts (Martial Arts Studies, 2), London/New York: Rowman and Littlefield, 2017, xxiii + 186 p.
by: Sixt Wetzler
Published: (2017)

“Your Kung Fu is very good, Master Fiore!” Asian and European fight books in comparison
by: Sixt Wetzler
Published: (2016)

PERSPECTIVES ON URBAN LONELINESS : City Paths Toward (Un)common Lifeworlds
by: Nina Goldman
Published: (2026)

Future Events as Backdoor Triggers: Investigating Temporal Vulnerabilities in LLMs
by: Price, Sara, et al.
Published: (2024)

No Difference in Pullout Strength Between a Bio‐inductive Implant and a Semitendinosus Tendon Graft in a Biomechanical Study of Medial Patellofemoral Ligament Repair Augmentation
by: Austin Wetzler, et al.
Published: (2024)

All Code, No Thought: Current Language Models Struggle to Reason in Ciphered Language
by: Guo, Shiyuan, et al.
Published: (2025)

The LLM Has Left The Chat: Evidence of Bail Preferences in Large Language Models
by: Ensign, Danielle, et al.
Published: (2025)

Uncovering Gaps in How Humans and LLMs Interpret Subjective Language
by: Jones, Erik, et al.
Published: (2025)

LocalBench: Benchmarking LLMs on County-Level Local Knowledge and Reasoning
by: Gao, Zihan, et al.
Published: (2025)

Same Answer, Different Representations: Hidden instability in VLMs
by: Wani, Farooq Ahmad, et al.
Published: (2026)

RippleBench: Capturing Ripple Effects Using Existing Knowledge Repositories
by: Rinberg, Roy, et al.
Published: (2025)