arXiv:2607. 03739v1 Announce Type: cross Abstract: We release a benchmark and failure-mode-aware evaluation framework for grounded QA under coordinated retrieval poisoning.
arXiv:2606. 12703v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) agents increasingly run with persistent memory that accumulates across user sessions.
The paper demonstrates that the way evaluation streams are assembled in streaming intrusion‑detection benchmarks—by interleaving, pooling, or replaying network captures—acts as an uncontrolled experimental variable that can significantly alter performance metrics. In the CICIDS2017 benchmark, reordering the same set of records under a fixed split changes the held‑out samples’ overlap, prevalence, and even reverses the ranking of two deterministic scorers. Similar effects are observed in the LITNET‑2020 benchmark, where pooling disjoint captures yields a single operating point that masks large variations in per‑capture prevalences, and minor changes in batch composition can shift reported AUC‑PR values by a few thousandths.
The paper investigates how allocating a query budget to structural depth rather than surface variation improves jailbreak success against the SAGE self‑check defense. By using a best‑of‑N approach over a code‑completion encoding, the authors achieve 67%, 22%, and 15% success rates on three open‑weight targets—far exceeding the 4.7% and 3.0% rates of single‑draw encoding and character‑search methods. The study demonstrates that depth of encoding and breadth of variation independently undermine transform and gate defenses, and that repeated sampling can inflate perceived robustness.
arXiv:2606. 02822v1 Announce Type: cross Abstract: Production LLM applications stack several defense families -- refusal-phrase filters, token-budget controls, model allowlists, rate limits, tool-registry authentication -- yet existing breach-and-attack-simulation (BAS) benchmarks report a single aggregate coverage number, hiding which family closes which threat.
arXiv:2609.00470v1 Announce Type: new Abstract: Retrieval-Augmented Generation (RAG) grounds large language models in external corpora, but implicit trust in retrieved documents creates a critical at...
arXiv:2607. 26574v1 Announce Type: cross Abstract: Safety classifiers ("guards") are the dominant black-box defense for vision-language models, yet they judge an input's surface form, not its meaning: a harmful request re-encoded as set theory, formal logic, a rare language, code, or an image of text slips past a guard that would block it in plain language -- the decode gap.
The paper introduces HackProbe, a black‑box monitoring tool that can be attached to any self‑evolving language model loop without accessing internal weights or activations. HackProbe uses a fixed‑distribution comparison core and a rotated fresh layer to detect reward hacking through four statistical tests, and it can immunize the model by selecting honest candidates from the proposal pool. Experiments on a controlled host with injected hacking channels show that HackProbe achieves higher AUROC and lower false‑positive rates than the strongest baseline, and its bandwidth‑limited reselection improves true capability under hacking more than it harms clean runs.
TriShieldRAG introduces a three‑layered defense for Retrieval‑Augmented Generation: an Ingest Guard that screens documents, a Retrieval Scorer that re‑ranks based on trust, and a Cross‑LLM Consensus that validates evidence across three models. Against the original PoisonedRAG attack on the 2.68M‑passage Natural Questions corpus, the framework reduces attack success from about 79% to 1%. However, adaptive attacks that only alter document formatting can bypass the Ingest Guard and still achieve high success rates, revealing limits of layered defenses that rely on the same retrieved evidence.
arXiv:2607. 01239v1 Announce Type: cross Abstract: Character-level perturbations bypass safety alignment in modern LLMs despite leaving prompts human-readable.
arXiv:2607. 26574v2 Announce Type: replace-cross Abstract: Safety classifiers ("guards") are the dominant black-box defense for vision-language models, yet a guard judges an input's surface form, not its meaning: a harmful request re-encoded as set theory, formal logic, a classical language, code, or text rendered inside an image slips past a guard that would block it in plain language - the decode gap.
arXiv:2608. 02678v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) systems are vulnerable to corpus poisoning: an attacker who inserts a crafted document into the retrieval corpus can steer the underlying large language model (LLM) toward an attacker-chosen wrong answer.
arXiv:2608.00745v3 Announce Type: replace Abstract: The literature on self-adapting malware is open-loop: adaptation is evaluated against static detectors in simulators, with fitness computed by expe...