arXiv:2508. 00042v2 Announce Type: replace-cross Abstract: Machine learning models deployed in non-stationary environments degrade silently, since as the input distribution drifts their accuracy decays without an error signal and without labels to reveal it.
arXiv:2607. 07146v1 Announce Type: new Abstract: The Internal Waves Service screens the Sentinel-1 Wave-mode archive for internal solitary waves, routing detections to experts whose adjudication time is the resource the effort exists to conserve.
arXiv:2603. 25670v3 Announce Type: replace Abstract: Safety monitoring is essential for Cyber-Physical Systems (CPSs).
arXiv:2602. 11619v2 Announce Type: replace Abstract: Running the same LLM agent on identical inputs yields 2.
arXiv:2607. 01025v1 Announce Type: cross Abstract: Radio-frequency (RF) sensing is a central modality for counter-unmanned-aerial-system (counter-UAS) defence because it exploits the control, telemetry, and video links between a drone and its operator.
arXiv:2602. 01515v2 Announce Type: replace-cross Abstract: Deploying learned control policies is risky because policies that appear robust in simulation can confidently enter out-of-distribution (OOD) states after Sim-to-Real transfer, causing silent failures and potential hardware damage.
arXiv:2608. 02786v1 Announce Type: new Abstract: AI systems can fail silently.
arXiv:2608. 05199v1 Announce Type: cross Abstract: Autonomous security agents operate as staged pipelines, such as classifying network traffic and then attributing attacks to a specific technique.
TERRA is an edge-AI framework for regional fire detection that performs task-oriented local inference and identity-free region-tagged reporting. It offers two reporting modes: TERRA-Score, which sends a quantized fire probability, and TERRA-Feature, which sends a compact learned representation for aggregation via a permutation-invariant DeepSets model. In tests on the EN54 indoor fire-alarm dataset, TERRA-Feature achieved a mean AUPRC of 0.990 with a 24‑bit payload, while TERRA-Score reached 0.987 with only 6 bits, both maintaining a low missed-detection rate of 1.81% at a false‑alarm rate of 1%.
The paper investigates adaptive UAV sampling for constructing radio environment maps (REMs) from non‑Gaussian IoT observations using Integrated Nested Laplace Approximation (INLA) to update a Matérn Gaussian Markov random field posterior. Waypoints are chosen via entropy pre‑screening and a Sherman–Morrison lookahead, and the study finds that widening the prior shadowing standard deviation consistently improves NMSE and credible‑interval coverage across multiple ray‑traced windows, while online width estimation from UAV counts does not recover this gain. Comparative analyses show that posterior‑informed sampling outperforms random and lawnmower strategies, but differences among informative acquisition rules remain unresolved.
arXiv:2609.26839v1 Announce Type: cross Abstract: Post-hoc probability calibration is usually evaluated under an optimistic assumption: the held-out calibration labels are clean. In many AI deploymen...
arXiv:2608. 02464v1 Announce Type: cross Abstract: LLM agents fail mid-episode -- they loop, cascade tool errors, drift off goal, fabricate results, or silently absorb corrupted content -- and the standard remedy, judging every step with a second LLM, costs more than the agent itself.