REVIEW 10 cited by
HiddenDetect: Detecting Jailbreak Attacks against Large Vision-Language Models via Monitoring Hidden States
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
The integration of additional modalities increases the susceptibility of large vision-language models (LVLMs) to safety risks, such as jailbreak attacks, compared to their language-only counterparts. While existing research primarily focuses on post-hoc alignment techniques, the underlying safety mechanisms within LVLMs remain largely unexplored. In this work , we investigate whether LVLMs inherently encode safety-relevant signals within their internal activations during inference. Our findings reveal that LVLMs exhibit distinct activation patterns when processing unsafe prompts, which can be leveraged to detect and mitigate adversarial inputs without requiring extensive fine-tuning. Building on this insight, we introduce HiddenDetect, a novel tuning-free framework that harnesses internal model activations to enhance safety. Experimental results show that {HiddenDetect} surpasses state-of-the-art methods in detecting jailbreak attacks against LVLMs. By utilizing intrinsic safety-aware patterns, our method provides an efficient and scalable solution for strengthening LVLM robustness against multimodal threats. Our code will be released publicly at https://github.com/leigest519/HiddenDetect.
Forward citations
Cited by 10 Pith papers
-
MMAligner: Safeguarding Multimodal Large Language Models through Representation Calibration
Safety failures in multimodal LLMs happen because unsafe image-plus-text inputs shift internal representations past a still-working refusal boundary, and a representation-calibration fine-tune restores refusal with un...
-
V-DEAL: Diagnosing Video Safety De-Calibration as an Understanding-Refusal Coupling Failure
Video LLMs understand harmful content but activate weaker refusal signals when the query is benign; prompt realignment reduces attack success from ~48% to ~1%.
-
Breaking Refusal in the First Half: A Mechanistic Study of the Prefill Jailbreak
Refusal under a prefill jailbreak is a shallow response-site computation: the harm representation stays intact, the failure lives in an early response window, and the dominant mechanism is passive autoregressive conditioning.
-
Adaptively Robust LLM Monitoring via Activation Watermarking
Activation Watermarking embeds a secret keyed direction in an LLM's hidden states so policy-violating responses can be detected by a cosine test, cutting adaptive-jailbreak evasion relative to guard models.
-
The First Differentiable Transfer-Based Algorithm for Discrete MicroLED Repair
The submission's abstract announces a differentiable microLED repair planner, yet its full text is the LoD jailbreak-detection paper (arXiv 2508.09201v4), leaving the claimed result without any derivation.
-
Learning Efficient Robotic Garment Manipulation with Standardization
APS-Net combines fling and pick-and-place actions to unfold and standardize garments, achieving better coverage, alignment, and real-world folding success than earlier methods.
-
Beyond Safe Answers: A Benchmark for Evaluating True Risk Awareness in Large Reasoning Models
A new benchmark shows that top reasoning models identify all relevant risks in under 40% of cases even when their final answers look safe.
-
USB: A Comprehensive and Unified Safety Evaluation Benchmark for Multimodal Large Language Models
USB-SafeBench is a unified MLLM safety benchmark with 61 risk categories, 4 modality combinations, and dual-language vulnerability and oversensitivity tests.
-
The Hidden Dimensions of LLM Alignment: A Multi-Dimensional Analysis of Orthogonal Safety Directions
Safety refusal in Llama 3.1 8B is governed by a dominant activation direction plus smaller interpretable directions, and removing prompt tokens that activate these secondary directions can bypass fine-tuned safety.
-
Language-Guided Long Horizon Manipulation with LLM-based Planning and Visual Perception
A robot folds cloth from spoken language by decomposing instructions with GPT-4o and grounding each step with a SigLIP2-based pick-and-place perception module.
Discussion (0). Sign in to comment.