LLM Safety From Within: Detecting Harmful Content with Internal Representations

• URL: http://arxiv.org/abs/2604.18519v1
• Date: Mon, 20 Apr 2026 17:17:07 GMT
• Title: LLM Safety From Within: Detecting Harmful Content with Internal Representations
• Abstract summary: We present SIREN, a lightweight guard model that harnesses safety-relevant features distributed across internal layers.<n>Our evaluation shows that SIREN substantially outperforms state-of-the-art open-source guard models.
• Score: 16.054258510018688
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: Guard models are widely used to detect harmful content in user prompts and LLM responses. However, state-of-the-art guard models rely solely on terminal-layer representations and overlook the rich safety-relevant features distributed across internal layers. We present SIREN, a lightweight guard model that harnesses these internal features. By identifying safety neurons via linear probing and combining them through an adaptive layer-weighted strategy, SIREN builds a harmfulness detector from LLM internals without modifying the underlying model. Our comprehensive evaluation shows that SIREN substantially outperforms state-of-the-art open-source guard models across multiple benchmarks while using 250 times fewer trainable parameters. Moreover, SIREN exhibits superior generalization to unseen benchmarks, naturally enables real-time streaming detection, and significantly improves inference efficiency compared to generative guard models. Overall, our results highlight LLM internal states as a promising foundation for practical, high-performance harmfulness detection.

Related papers

• Robust Multimodal Safety via Conditional Decoding [52.92816441364308]
  Multimodal large-language models (MLLMs) often experience degraded safety alignment when harmful queries exploit cross-modal interactions.<n>We propose a simple conditional decoding strategy, CASA, that utilizes internal representations of MLLMs to predict a binary safety token before response generation.
  arXiv  Detail & Related papers  (2026-03-31T23:19:50Z)
• Finding and Reactivating Post-Trained LLMs' Hidden Safety Mechanisms [43.21675006313594]
  This paper investigates the underlying cause of safety degradation in large language models (LLMs)<n>We propose a lightweight and cost-effective solution called SafeReAct that restores the suppressed safety behaviors by aligning with LoRA adapters.<n> Experiments on four state-of-the-art LRMs show that our method significantly improves safety on harmful prompts without compromising reasoning performance.
  arXiv  Detail & Related papers  (2026-03-10T06:56:23Z)
• Enhancing Safety of Large Language Models via Embedding Space Separation [24.821172423424866]
  Large language models (LLMs) have achieved impressive capabilities, yet ensuring their safety against harmful prompts remains a critical challenge.<n>Recent work has revealed that the latent representations (embeddings) of harmful and safe queries in LLMs typically exhibit linear separability.<n>Motivated by this observation, we propose a representation-level fine-tuning approach, named Embedding Space Separation (ES2)
  arXiv  Detail & Related papers  (2026-03-01T07:22:38Z)
• Taxonomy-Adaptive Moderation Model with Robust Guardrails for Large Language Models [3.710103086278309]
  Large Language Models (LLMs) are typically aligned for safety during the post-training phase.<n>They may still generate inappropriate outputs that could potentially pose risks to users.<n>This challenge underscores the need for robust safeguards that operate across both model inputs and outputs.
  arXiv  Detail & Related papers  (2025-12-05T00:43:55Z)
• DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models [50.21378052667732]
  We conduct an in-depth analysis of dLLM vulnerabilities to jailbreak attacks across two distinct dimensions: intra-step and inter-step dynamics.<n>We propose DiffuGuard, a training-free defense framework that addresses vulnerabilities through a dual-stage approach.
  arXiv  Detail & Related papers  (2025-09-29T05:17:10Z)
• ReGA: Representation-Guided Abstraction for Model-based Safeguarding of LLMs [0.9285458070502282]
  Large Language Models (LLMs) have achieved significant success in various tasks, yet concerns about their safety and security have emerged.<n>To analyze and monitor machine learning models, model-based analysis has demonstrated notable potential in stateful deep neural networks.<n>We propose ReGA, a model-based analysis framework with representation-guided abstraction, to safeguard LLMs against harmful prompts and generations.
  arXiv  Detail & Related papers  (2025-06-02T15:17:38Z)
• LookAhead Tuning: Safer Language Models via Partial Answer Previews [62.529794567687354]
  Fine-tuning enables large language models to adapt to specific domains, but often compromises their previously established safety alignment.<n>We introduce LookAhead Tuning, a lightweight and effective data-driven approach that preserves safety during fine-tuning.
  arXiv  Detail & Related papers  (2025-03-24T18:11:42Z)
• SafeSwitch: Steering Unsafe LLM Behavior via Internal Activation Signals [51.49737867797442]
  Large language models (LLMs) exhibit exceptional capabilities across various tasks but also pose risks by generating harmful content.<n>We show that LLMs can similarly perform internal assessments about safety in their internal states.<n>We propose SafeSwitch, a framework that regulates unsafe outputs by utilizing the prober-based internal state monitor.
  arXiv  Detail & Related papers  (2025-02-03T04:23:33Z)
• ShieldGemma: Generative AI Content Moderation Based on Gemma [49.91147965876678]
  ShieldGemma is a suite of safety content moderation models built upon Gemma2. Models provide robust, state-of-the-art predictions of safety risks across key harm types.
  arXiv  Detail & Related papers  (2024-07-31T17:48:14Z)
• On Prompt-Driven Safeguarding for Large Language Models [172.13943777203377]
  We find that in the representation space, the input queries are typically moved by safety prompts in a "higher-refusal" direction. Inspired by these findings, we propose a method for safety prompt optimization, namely DRO. Treating a safety prompt as continuous, trainable embeddings, DRO learns to move the queries' representations along or opposite the refusal direction, depending on their harmfulness.
  arXiv  Detail & Related papers  (2024-01-31T17:28:24Z)

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.