SABER: Uncovering Vulnerabilities in Safety Alignment via Cross-Layer Residual Connection

• URL: http://arxiv.org/abs/2509.16060v1
• Date: Fri, 19 Sep 2025 15:10:19 GMT
• Title: SABER: Uncovering Vulnerabilities in Safety Alignment via Cross-Layer Residual Connection
• Authors: Maithili Joshi, Palash Nandi, Tanmoy Chakraborty,
• Abstract summary: Large Language Models (LLMs) with safe-alignment training are powerful instruments with robust language comprehension capabilities.<n>We introduce a novel white-box jailbreak method, SABER, which connects two intermediate layers $s$ and $e$ such that $s e$, through a residual connection.<n>Our approach achieves a 51% improvement over the best-performing baseline on the HarmBench test set.
• Score: 16.38885847999291
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large Language Models (LLMs) with safe-alignment training are powerful instruments with robust language comprehension capabilities. These models typically undergo meticulous alignment procedures involving human feedback to ensure the acceptance of safe inputs while rejecting harmful or unsafe ones. However, despite their massive scale and alignment efforts, LLMs remain vulnerable to jailbreak attacks, where malicious users manipulate the model to produce harmful outputs that it was explicitly trained to avoid. In this study, we find that the safety mechanisms in LLMs are predominantly embedded in the middle-to-late layers. Building on this insight, we introduce a novel white-box jailbreak method, SABER (Safety Alignment Bypass via Extra Residuals), which connects two intermediate layers $s$ and $e$ such that $s < e$, through a residual connection. Our approach achieves a 51% improvement over the best-performing baseline on the HarmBench test set. Furthermore, SABER induces only a marginal shift in perplexity when evaluated on the HarmBench validation set. The source code is publicly available at https://github.com/PalGitts/SABER.

Related papers

• Attributing and Exploiting Safety Vectors through Global Optimization in Large Language Models [50.91504059485288]
  We propose a framework that identifies safety-critical attention heads through global optimization over all heads simultaneously.<n>We develop a novel inference-time white-box jailbreak method that exploits the identified safety vectors through activation repatching.
  arXiv  Detail & Related papers  (2026-01-22T09:32:43Z)
• The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs [39.85609149662187]
  We present DIJA, the first systematic study and jailbreak attack framework that exploits unique safety weaknesses of dLLMs.<n>Our proposed DIJA constructs adversarial interleaved mask-text prompts that exploit the text generation mechanisms of dLLMs.<n>Our findings underscore the urgent need for rethinking safety alignment in this emerging class of language models.
  arXiv  Detail & Related papers  (2025-07-15T08:44:46Z)
• Why Not Act on What You Know? Unleashing Safety Potential of LLMs via Self-Aware Guard Enhancement [48.50995874445193]
  Large Language Models (LLMs) have shown impressive capabilities across various tasks but remain vulnerable to meticulously crafted jailbreak attacks.<n>We propose SAGE (Self-Aware Guard Enhancement), a training-free defense strategy designed to align LLMs' strong safety discrimination performance with their relatively weaker safety generation ability.
  arXiv  Detail & Related papers  (2025-05-17T15:54:52Z)
• Playing the Fool: Jailbreaking LLMs and Multimodal LLMs with Out-of-Distribution Strategy [31.03584769307822]
  We propose JOOD, a new Jailbreak framework via OOD-ifying inputs beyond the safety alignment.<n>Experiments across diverse jailbreak scenarios demonstrate that JOOD effectively jailbreaks recent proprietary LLMs and MLLMs.
  arXiv  Detail & Related papers  (2025-03-26T01:25:24Z)
• Improving LLM Safety Alignment with Dual-Objective Optimization [65.41451412400609]
  Existing training-time safety alignment techniques for large language models (LLMs) remain vulnerable to jailbreak attacks.<n>We propose an improved safety alignment that disentangles DPO objectives into two components: (1) robust refusal training, which encourages refusal even when partial unsafe generations are produced, and (2) targeted unlearning of harmful knowledge.
  arXiv  Detail & Related papers  (2025-03-05T18:01:05Z)
• xJailbreak: Representation Space Guided Reinforcement Learning for Interpretable LLM Jailbreaking [32.89084809038529]
  Black-box jailbreak is an attack where crafted prompts bypass safety mechanisms in large language models.<n>We propose a novel black-box jailbreak method leveraging reinforcement learning (RL)<n>We introduce a comprehensive jailbreak evaluation framework incorporating keywords, intent matching, and answer validation to provide a more rigorous and holistic assessment of jailbreak success.
  arXiv  Detail & Related papers  (2025-01-28T06:07:58Z)
• LIAR: Leveraging Inference Time Alignment (Best-of-N) to Jailbreak LLMs in Seconds [98.20826635707341]
  Jailbreak attacks expose vulnerabilities in safety-aligned LLMs by eliciting harmful outputs through carefully crafted prompts.<n>We frame jailbreaks as inference-time misalignment and introduce LIAR, a fast, black-box, best-of-$N$ sampling attack requiring no training.<n>We also introduce a theoretical "safety net against jailbreaks" metric to quantify safety alignment strength and derive suboptimality bounds.
  arXiv  Detail & Related papers  (2024-12-06T18:02:59Z)
• How Alignment and Jailbreak Work: Explain LLM Safety through Intermediate Hidden States [65.45603614354329]
  Large language models (LLMs) rely on safety alignment to avoid responding to malicious user inputs. Jailbreak can circumvent safety guardrails, resulting in LLMs generating harmful content. We employ weak classifiers to explain LLM safety through the intermediate hidden states.
  arXiv  Detail & Related papers  (2024-06-09T05:04:37Z)
• Emulated Disalignment: Safety Alignment for Large Language Models May Backfire! [65.06450319194454]
  Large language models (LLMs) undergo safety alignment to ensure safe conversations with humans. This paper introduces a training-free attack method capable of reversing safety alignment. We name this method emulated disalignment (ED) because sampling from this contrastive distribution provably emulates the result of fine-tuning to minimize a safety reward.
  arXiv  Detail & Related papers  (2024-02-19T18:16:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.