arXiv:2606. 29646v1 Announce Type: cross Abstract: Sleeper agents are the canonical model organism of deception: models trained to behave normally but to emit an unsafe behaviour on a specific trigger.
arXiv:2607. 26849v1 Announce Type: cross Abstract: As large language models (LLMs) are deployed in high-stakes domains, adversaries may poison training data to implant backdoors: hidden triggers that covertly manipulate model behavior at inference time.
As large language models (LLMs) are deployed in high-stakes domains, adversaries may poison training data to implant backdoors: hidden triggers that covertly manipulate model behavior at inference time. We ask whether a defender can recover such a trigger under realistic affordances, namely white-box access to the weights and knowledge of the behavior of concern, but no training data, no trusted reference model, no knowledge of the trigger, and no certainty that the model is poisoned.
The paper investigates whether the hidden activations of large language models (LLMs) contain signals about the vulnerability of C/C++ code when the code is provided as context. By extracting prefill token activations from four LLMs and training small MLP probes, the authors achieve an average F1 score of 41.7% across four benchmarks, with the best probe matching state‑of‑the‑art fine‑tuned classifiers on the Devign dataset. The results suggest that a coding LLM’s internal representation can inform vulnerability detection, opening the door to lightweight, model‑native screening methods.
NeuronFuzz is a white‑box fuzzing framework that uses internal safety neurons of large language models as continuous feedback for safety evaluation, eliminating the need to generate full model responses during testing. It constructs a SafetyOracle that converts neuron activations into a differentiable safety alarm score, enabling gradient‑based identification of sensitive template positions and fluent, context‑compatible prompt mutations. Evaluated on 21 text and multimodal models, NeuronFuzz achieves a 76‑100% jailbreak discovery rate on five white‑box source models and demonstrates strong zero‑shot transfer to open‑weight and proprietary targets.
arXiv:2605. 00994v2 Announce Type: replace-cross Abstract: Finetuning can significantly modify the behavior of large language models, including introducing harmful or unsafe behaviors.
arXiv:2608.31105v1 Announce Type: new Abstract: Users of a deployed language model routinely encounter behaviours that testing almost never surfaces, since deployment puts the model through orders of...
The paper investigates whether adding a short classification instruction after a user’s prompt improves the ability of activation probes to detect malicious inputs in large language models. Across 13 safety benchmarks and three open‑weight model families, a classification suffix consistently boosts out‑of‑distribution detection (up to ~4 AUC points) compared to no suffix, and the benefit transfers to multi‑position pooling probes used in production. The improvement stems from the classification format itself rather than the specific content of the instruction, though the optimal suffix varies with the model and readout type.
The paper introduces Speculative Probing, a method that repurposes the speculative‑decoding module of large language models for real‑time classification tasks. By appending a trained soft prompt to the target sequence, the approach leverages the already‑cached KV store during inference, adding negligible overhead while achieving higher accuracy than traditional hidden‑state probes. Experiments on four classification tasks across multiple models show that these lightweight probes outperform zero‑shot GPT‑5.4‑mini and rival or surpass specialized 8B safety classifiers without running a full LLM.
arXiv:2607. 19894v1 Announce Type: cross Abstract: Large language models (LLMs) are vulnerable to backdoor attacks, where hidden triggers induce malicious outputs.
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. 24893v1 Announce Type: cross Abstract: Multi-agent LLM systems can be attacked by a payload that no single agent ever holds in full: a poisoned tool hides encrypted fragments in its observations, spreads them across several agents, and an external step reassembles and executes them after the run.
arXiv:2607. 11475v1 Announce Type: new Abstract: Safety alignment in large language models can be fragile under fine-tuning, as even benign task adaptation may increase harmful compliance.