arXiv:2510. 05709v2 Announce Type: replace-cross Abstract: LLM benchmarking metrics often misstate performance and uncertainty as they rely on two assumptions that frequently do not hold in practice: (i) a sufficient number of evaluations are available for classical inference, and (ii) test prompts are independent.
arXiv:2607. 03075v1 Announce Type: new Abstract: Safety-critical applications require classifiers that are both robust and reliable.
arXiv:2602. 09161v2 Announce Type: replace-cross Abstract: Simulation-based inference (SBI) enables amortized Bayesian inference by first training a neural posterior estimator (NPE) on prior-simulator pairs, typically through low-dimensional summary statistics, which can then be cheaply reused for fast inference by querying it on new test observations.
arXiv:2510. 09288v2 Announce Type: replace-cross Abstract: The vulnerability of machine learning models to adversarial attacks remains a critical societal security challenge.
The paper introduces CLEAR, a lightweight, task‑agnostic post‑hoc method that enhances evidential robustness in deep learning models without retraining. CLEAR uses held‑out calibration data to map the geometry of the model’s latent space, then generates perturbation views at inference to detect latent conflict. When high conflict is found, CLEAR selectively reduces evidential strength while preserving evidence for latent‑consistent inputs, achieving significant improvements in OOD and adversarial AUROC on ImageNet→CUB and running much faster than competing methods.
arXiv:2604. 23099v2 Announce Type: replace-cross Abstract: Evaluating generative AI models is increasingly resource-intensive due to slow inference, expensive raters, and a rapidly growing landscape of models and benchmarks.
The paper introduces techniques for measuring the robustness of predictions made by two generative classifiers—naive Bayes classifiers and generative forests—whose underlying models are probabilistic graphical models. Robustness is defined as the degree to which the classifier’s distribution can be perturbed without altering its prediction, with perturbations explored via epsilon‑contamination, total variation distance, and chi‑squared divergence neighborhoods. Experiments on benchmark datasets show that the computed robustness values can serve as indicators of prediction trustworthiness and are compared against other existing indicators.
arXiv:2606. 09878v1 Announce Type: new Abstract: Standard benchmarks report aggregate accuracy, but practitioners need to know which specific capabilities a model lacks.
GRAPE is a two‑stage Bayesian optimization framework that first refines the local gradient posterior using a closed‑form acquisition function and then selects update directions by maximizing expected decrease conditioned on descent. The authors prove that the refinement stage monotonically reduces local uncertainty and that the progress‑aware direction converges to true steepest descent as the posterior sharpens. Empirical results show GRAPE achieves a 5.4× speedup on black‑box adversarial attacks and reduces final average regret by 3.8 log‑units on large language model prompt‑optimization tasks.
arXiv:2607. 27023v1 Announce Type: new Abstract: Evaluating large generative models across benchmarks is time-consuming and computationally expensive.
arXiv:2607. 08590v1 Announce Type: new Abstract: Scientific experiments are often designed to maximize information gain, yet in many applications the primary objective is to support reliable downstream decision-making.
arXiv:2410. 07719v4 Announce Type: replace Abstract: Despite being widely adopted as a canonical framework for learning robust models, adversarial training suffers from robust overfitting.
arXiv:2511.10661v2 Announce Type: replace Abstract: It is increasingly important to evaluate the characteristics of systems based on large language models (LLMs). Evaluations in this context often re...