Fine-tuning updates frequently stale activation monitors for language model safety while quantization does not, with degradation predictable and repairable via label-free realignment.
Safety layers of aligned large language models: The key to llm security
11 Pith papers cite this work. Polarity classification is still indexing.
11
Pith papers citing it
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
Sequential LLM defense deployment leads to risk exacerbation in 38.9% of cases due to anti-aligned updates in shared critical layers, addressed by conflict-guided layer freezing.
A truly benign DPO attack using 10 harmless preference pairs jailbreaks frontier LLMs by suppressing refusal behavior, achieving up to 81.73% attack success rate on GPT-4.1-nano at low cost.
SIREN identifies safety neurons via linear probing on internal LLM layers and combines them with adaptive weighting to detect harm, outperforming prior guard models with 250x fewer parameters.
Coupled constraints on weight updates in a safety subspace and regularization of SAE-identified safety features preserve LLM refusal behaviors during fine-tuning better than weight-only or activation-only methods.
Causal mediation analysis shows harmful LLM outputs arise in late layers from MLP failures and gating neurons, with early layers handling harm context detection and signal propagation.
LEG is a compact model that jointly classifies unsafe prompts and explains its decisions using synthetic training data and a custom uncertainty-weighted loss.
Jailbreak attacks suppress Adversarially Compromised Heads in early layers but leave Safety-Aligned Heads active in mid-layers, producing robust harmful features usable for competitive training-free detection.
SALLIE detects jailbreaks in text and vision-language models by extracting residual stream activations, scoring maliciousness per layer with k-NN, and ensembling predictions, outperforming baselines on multiple datasets.
Survey of harmful fine-tuning attacks on LLMs, their variants, defense strategies, mechanical analysis, and evaluation methodologies.
citing papers explorer
-
Defenses at Odds: Measuring and Explaining Defense Conflicts in Large Language Models
Sequential LLM defense deployment leads to risk exacerbation in 38.9% of cases due to anti-aligned updates in shared critical layers, addressed by conflict-guided layer freezing.
-
Few-Shot Truly Benign DPO Attack for Jailbreaking LLMs
A truly benign DPO attack using 10 harmless preference pairs jailbreaks frontier LLMs by suppressing refusal behavior, achieving up to 81.73% attack success rate on GPT-4.1-nano at low cost.
-
Robust Harmful Features Under Jailbreak Attacks: Mechanistic Evidence from Attention Head Specialization in Large Language Models
Jailbreak attacks suppress Adversarially Compromised Heads in early layers but leave Safety-Aligned Heads active in mid-layers, producing robust harmful features usable for competitive training-free detection.
-
SALLIE: Safeguarding Against Latent Language & Image Exploits
SALLIE detects jailbreaks in text and vision-language models by extracting residual stream activations, scoring maliciousness per layer with k-NN, and ensembling predictions, outperforming baselines on multiple datasets.
-
Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Survey
Survey of harmful fine-tuning attacks on LLMs, their variants, defense strategies, mechanical analysis, and evaluation methodologies.