Cover Image

Safety-Aligned Weights Are Not Enough: Refusal-Teacher-Guided Finetuning Enhances Safety and Downstream Performance under Harmful Finetuning Attacks

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ham, Seokil, Choi, Yubin, Yang, Yujin, Cho, Seungju, Kim, Younghun, Kim, Changick
Format:	Preprint
Published:	2025
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Locking Down the Finetuned LLMs Safety
by: Zhu, Minjun, et al.
Published: (2024)

Parameter Efficient Mamba Tuning via Projector-targeted Diagonal-centric Linear Transformation
by: Ham, Seokil, et al.
Published: (2024)

Enhancing Robustness in Incremental Learning with Adversarial Training
by: Cho, Seungju, et al.
Published: (2023)

Should LLM Safety Be More Than Refusing Harmful Instructions?
by: Maskey, Utsav, et al.
Published: (2025)

Large Language Models to Diffusion Finetuning
by: Cetin, Edoardo, et al.
Published: (2025)

WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs
by: Han, Seungju, et al.
Published: (2024)

Refusal Direction is Universal Across Safety-Aligned Languages
by: Wang, Xinpeng, et al.
Published: (2025)

From Threat to Tool: Leveraging Refusal-Aware Injection Attacks for Safety Alignment
by: Chae, Kyubyung, et al.
Published: (2025)

Long-tailed Adversarial Training with Self-Distillation
by: Cho, Seungju, et al.
Published: (2025)

Indirect Gradient Matching for Adversarial Robust Distillation
by: Lee, Hongsin, et al.
Published: (2023)

Representation Noising: A Defence Mechanism Against Harmful Finetuning
by: Rosati, Domenic, et al.
Published: (2024)

Time is Encoded in the Weights of Finetuned Language Models
by: Nylund, Kai, et al.
Published: (2023)

TS-Align: A Teacher-Student Collaborative Framework for Scalable Iterative Finetuning of Large Language Models
by: Zhang, Chen, et al.
Published: (2024)

Finetune-Informed Pretraining Boosts Downstream Performance
by: Faysal, Atik, et al.
Published: (2026)

Diffusion Model Patching via Mixture-of-Prompts
by: Ham, Seokil, et al.
Published: (2024)

Switch Diffusion Transformer: Synergizing Denoising Tasks with Sparse Mixture-of-Experts
by: Park, Byeongjun, et al.
Published: (2024)

ELITE: Enhanced Language-Image Toxicity Evaluation for Safety
by: Lee, Wonjun, et al.
Published: (2025)

Safety is Not Only About Refusal: Reasoning-Enhanced Fine-tuning for Interpretable LLM Safety
by: Zhang, Yuyou, et al.
Published: (2025)

RICoTA: Red-teaming of In-the-wild Conversation with Test Attempts
by: Choi, Eujeong, et al.
Published: (2025)

When Scaling Meets LLM Finetuning: The Effect of Data, Model and Finetuning Method
by: Zhang, Biao, et al.
Published: (2024)

Expected Harm: Rethinking Safety Evaluation of (Mis)Aligned LLMs
by: Chen, Yen-Shan, et al.
Published: (2026)

The Hallucination Tax of Reinforcement Finetuning
by: Song, Linxin, et al.
Published: (2025)

Detecting AI-Generated Videos with Spiking Neural Networks
by: Jang, Minsuk, et al.
Published: (2026)

Synthesizing Privacy-Preserving Text Data via Finetuning without Finetuning Billion-Scale LLMs
by: Tan, Bowen, et al.
Published: (2025)

Order Independence With Finetuning
by: Brown, Katrina, et al.
Published: (2025)

Sparse Memory Finetuning as a Low-Forgetting Alternative to LoRA and Full Finetuning
by: Gupta, Prakhar, et al.
Published: (2026)

Representation Bending for Large Language Model Safety
by: Yousefpour, Ashkan, et al.
Published: (2025)

LLMs Encode Harmfulness and Refusal Separately
by: Zhao, Jiachen, et al.
Published: (2025)

Improving Sparse Memory Finetuning
by: Goyal, Satyam, et al.
Published: (2026)

Predicting Emergent Capabilities by Finetuning
by: Snell, Charlie, et al.
Published: (2024)

Whisper Finetuning on Nepali Language
by: Rijal, Sanjay, et al.
Published: (2024)

High Accuracy, Less Talk (HALT): Reliable LLMs through Capability-Aligned Finetuning
by: Franzmeyer, Tim, et al.
Published: (2025)

GradShield: Alignment Preserving Finetuning
by: Hu, Zhanhao, et al.
Published: (2026)

Understanding the Effects of Domain Finetuning on LLMs
by: Tanwar, Eshaan, et al.
Published: (2025)

Finetuning LLMs for Comparative Assessment Tasks
by: Raina, Vatsal, et al.
Published: (2024)

Surgical Refusal Ablation: Disentangling Safety from Intelligence via Concept-Guided Spectral Cleaning
by: Cristofano, Tony
Published: (2026)

Think Before Refusal : Triggering Safety Reflection in LLMs to Mitigate False Refusal Behavior
by: Si, Shengyun, et al.
Published: (2025)

ReasAlign: Reasoning Enhanced Safety Alignment against Prompt Injection Attack
by: Li, Hao, et al.
Published: (2026)

KnowLA: Enhancing Parameter-efficient Finetuning with Knowledgeable Adaptation
by: Luo, Xindi, et al.
Published: (2024)

Enhancing Talent Employment Insights Through Feature Extraction with LLM Finetuning
by: Thakrar, Karishma, et al.
Published: (2025)