arXiv:2605. 00994v2 Announce Type: replace-cross Abstract: Finetuning can significantly modify the behavior of large language models, including introducing harmful or unsafe behaviors.
The paper investigates how trigger-based backdoors function in large language models by using sparse autoencoders (SAEs) to identify feature directions across layers and transformer components. In a controlled experiment, 1B and 8B models were made to continue English prompts in French or German when presented with fixed trigger sequences. The study finds that different SAE feature directions correspond to trigger detection, residual-stream propagation, and language tracking, with residual-stream features being most effective for controlling the backdoor behavior.
Backdoor attacks in Large Language Models (LLMs) are a growing security concern, where models can generate adversary-chosen content. Existing defenses target backdoors one at a time and typically require knowledge of the trigger, leaving the defender at a structural disadvantage when unknown backdoors may exist in a model.
UniGuardian is a training‑free detector for large language models that jointly identifies prompt injection, backdoor, and adversarial attacks—collectively called Prompt Trigger Attacks (PTA). It measures how structured prompt perturbations shift the model’s output distribution and uses a single‑forward strategy to detect attacks while generating text in a shared batched forward pass. Experiments show that UniGuardian accurately and efficiently identifies trigger‑activated prompts in LLMs.
The paper identifies a vulnerability in large language models where harmful intent can be hidden within benign narratives, a phenomenon termed Semantic Camouflage. By examining latent activation patterns across several small language model families, the authors discover an "Intent Horizon"—a layer depth where harmful intent representations collapse. They propose Latent Intent Verification (LIV), a lightweight probing defense that detects harmful intent in early layers and outperforms existing guardrails on the PKU-SafeRLHF dataset.
arXiv:2606. 07963v1 Announce Type: new Abstract: Backdoor attacks in large language models (LLMs) are often treated as isolated trigger-response failures, motivating defenses tailored to specific triggers or behaviors.
Large language models (LLMs) are vulnerable to backdoor attacks, where hidden triggers induce malicious outputs. Existing defenses generally fall into inference-time detection or training-time mitigation, but face two key limitations.
arXiv:2607. 19894v1 Announce Type: cross Abstract: Large language models (LLMs) are vulnerable to backdoor attacks, where hidden triggers induce malicious outputs.
arXiv:2606. 02995v1 Announce Type: cross Abstract: Large language models remain vulnerable to jailbreak backdoor attacks, where adversaries poison safety alignment data to embed hidden triggers that bypass safety mechanisms.
arXiv:2607. 01208v1 Announce Type: cross Abstract: Language models deployed in high-stakes roles can potentially favor certain entities, brands, or viewpoints, steering user decisions at scale.
The paper introduces Attnlocate, a runtime framework that localizes behavior‑guiding instructions within the attention matrix of large language model agents. By treating this localization as an object detection task, Attnlocate uses a multi‑head, multi‑layer attention aggregation scheme and a 1‑D U‑Net to identify spans that influence tool‑calling decisions. The system then adjudicates potential malicious invocations based on the authority of the source, achieving high detection metrics across diverse LLM families and demonstrating transferability to unseen models.
Large language models (LLMs) are rigorously aligned to refuse harmful requests, a process that inherently cultivates a latent capacity to evaluate and recognize unsafe content. In this work, we reveal that this advanced safety awareness inadvertently introduces a fatal vulnerability.