# SALLIE: Generation-Free Hidden-State Detection of Jailbreaks and Prompt Injections Across Text and Vision

arXiv: https://arxiv.org/abs/2604.06247 · HTML: https://mlsanity.com/paper/2604.06247/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 29 of its 49 references that are in the map.
- First posted on arXiv 6 Apr 2026.

## Placed in

- Multimodal LLMs / vision-language models / [MLLM jailbreaks and adversarial robustness](https://mlsanity.com/mllm/mllm-jailbreaks-and-adversarial-robustness/)
- LLM safety and security / [Vision-language and text-to-image safety](https://mlsanity.com/safety/vision-language-and-text-to-image-safety/)

## Builds on (most cited first, 29 of 29)

- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI](https://mlsanity.com/paper/2311.16502/) (2023)
- [Towards VQA Models That Can Read](https://mlsanity.com/paper/1904.08920/) (2019)
- [Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection](https://mlsanity.com/paper/2302.12173/) (2023)
- [MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities](https://mlsanity.com/paper/2308.02490/) (2023)
- [OK-VQA: A Visual Question Answering Benchmark Requiring External Knowledge](https://mlsanity.com/paper/1906.00067/) (2019)
- [Ignore Previous Prompt: Attack Techniques For Language Models](https://mlsanity.com/paper/2211.09527/) (2022)
- [Baseline Defenses for Adversarial Attacks Against Aligned Language Models](https://mlsanity.com/paper/2309.00614/) (2023)
- [MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmark](https://mlsanity.com/paper/2409.02813/) (2024)
- [Mitigating Hallucination in Large Multi-Modal Models via Robust Instruction Tuning](https://mlsanity.com/paper/2306.14565/) (2023)
- [Detecting Language Model Attacks with Perplexity](https://mlsanity.com/paper/2308.14132/) (2023)
- [FigStep: Jailbreaking Large Vision-Language Models via Typographic Visual Prompts](https://mlsanity.com/paper/2311.05608/) (2023)
- [Visual Adversarial Examples Jailbreak Aligned Large Language Models](https://mlsanity.com/paper/2306.13213/) (2023)
- [LLM Self Defense: By Self Examination, LLMs Know They Are Being Tricked](https://mlsanity.com/paper/2308.07308/) (2023)
- [SmolVLM: Redefining small and efficient multimodal models](https://mlsanity.com/paper/2504.05299/) (2025)
- [JailBreakV: A Benchmark for Assessing the Robustness of MultiModal Large Language Models against Jailbreak Attacks](https://mlsanity.com/paper/2404.03027/) (2024)
- [MMC: Advancing Multimodal Chart Understanding with Large-scale Instruction Tuning](https://mlsanity.com/paper/2311.10774/) (2023)
- [Tensor Trust: Interpretable Prompt Injection Attacks from an Online Game](https://mlsanity.com/paper/2311.01011/) (2023)
- [Safety Layers in Aligned Large Language Models: The Key to LLM Security](https://mlsanity.com/paper/2408.17003/) (2024)
- [How Alignment and Jailbreak Work: Explain LLM Safety through Intermediate Hidden States](https://mlsanity.com/paper/2406.05644/) (2024)
- [Towards Understanding Jailbreak Attacks in LLMs: A Representation Space Analysis](https://mlsanity.com/paper/2406.10794/) (2024)
- [VisualWebInstruct: Scaling up Multimodal Instruction Data through Web Search](https://mlsanity.com/paper/2503.10582/) (2025)
- [HiddenDetect: Detecting Jailbreak Attacks against Large Vision-Language Models via Monitoring Hidden States](https://mlsanity.com/paper/2502.14744/) (2025)
- [JailbreakLens: Interpreting Jailbreak Mechanism in the Lens of Representation and Circuit](https://mlsanity.com/paper/2411.11114/) (2024)
- [PIShield: Detecting Prompt Injection Attacks via Intrinsic LLM Features](https://mlsanity.com/paper/2510.14005/) (2025)
- [VLM-Guard: Safeguarding Vision-Language Models via Fulfilling Safety Alignment Gap](https://mlsanity.com/paper/2502.10486/) (2025)

## Abstract (arXiv)

Large Language Models (LLMs) and Vision-Language Models (VLMs) are vulnerable to jailbreaks and prompt injections delivered through text or images. Existing defenses often narrow threat coverage or add inference cost through input transformations, repeated model queries, generated responses, or threat-specific detectors. We present SALLIE (Single-Pass Activation Lookup for Layerwise Input Evaluation), a white-box, generation-free detector with a common architecture for textual and visual jailbreaks and prompt injections. Using one backbone forward pass, SALLIE extracts last-token residual-stream representations, applies layer-wise k-nearest-neighbor probes, and averages their scores across a contiguous layer range, without modifying the backbone or generating a response. To distinguish reusable detector configurations from deployment-specific tuning, we compare shared, threshold-only, and fully calibrated regimes across backbones and modalities. We evaluate SALLIE on compact open-weight VLMs (Gemma-3-4b-it, Phi-3.5-vision-instruct, SmolVLM2-2.2B-Instruct) using heterogeneous held-out test data spanning both threat families and modalities. Fully calibrated SALLIE-Phi achieves an aggregate F1 of 0.87 and higher aggregate balanced accuracy than RCS-KCD on all three backbones under a matched-calibration protocol. On Phi-3.5 visual inputs, threshold-only calibration yields an F1 of 0.99 with zero observed false positives, versus 0.54 for zero-shot Gemini-2.5-Flash-Lite. On Gemma-4-E2B-it, a backbone held out from representation-hyperparameter selection, the same procedure raises F1 from 0.62 to 0.79 on text and 0.68 to 0.98 on visual inputs. Representation hyperparameters can thus transfer across some backbones, while operating points remain backbone-, modality-, and distribution-dependent. We further report FPR/FNR tradeoffs, latency, and representation ablations.