Chain-of-thought (CoT) reasoning with self-consistency improves performance by aggregating multiple sampled reasoning paths. In this setting, correctness is no longer tied to a single reasoning trace but to the aggregation rule over a pool of candidate paths, making aggregation uncertainty the central challenge. This issue is critical where confidently incorrect answers are far more costly than abstentions. We introduce a conformal procedure for CoT reasoning that directly addresses aggregation uncertainty. Our approach replaces majority voting with weighted score aggregation over reasoning paths and calibrates an abstention rule using conformal risk control. This approach leads to finite-sample guarantees on the confident-error rate--the probability that the system answers and is wrong. We further identify score separability as the key condition under which abstention provably improves selective accuracy, and derive closed-form expressions that predict accuracy gains from calibration data alone. The method is fully inference-time, and requires no retraining. Across four benchmarks, four open-source models, and three score classes, realized confident-error rates are consistent with the prescribed targets up to calibration-split and test-set variability. Our method achieves $90.1\%$ selective accuracy on GSM8K by abstaining on less than $5\%$ of problems, compared with $82\%$ accuracy under majority-voting baseline.

The rise of misinformation underscores the need for scalable and reliable fact-checking solutions. Large language models (LLMs) hold promise in automating fact verification, yet their effectiveness across global contexts remains uncertain. We systematically evaluate nine established LLMs across multiple categories (open/closed-source, multiple sizes, diverse architectures, reasoning-based) using 5,000 claims previously assessed by 174 professional fact-checking organizations across 47 languages. Our methodology tests model generalizability on claims postdating training cutoffs and four prompting strategies mirroring both citizen and professional fact-checker interactions, with over 240,000 human annotations as ground truth. Findings reveal a concerning pattern resembling the Dunning-Kruger effect: smaller, accessible models show high confidence despite lower accuracy, while larger models demonstrate higher accuracy but lower confidence. This risks systemic bias in information verification, as resource-constrained organizations typically use smaller models. Performance gaps are most pronounced for non-English languages and claims originating from the Global South, threatening to widen existing information inequalities. These results establish a multilingual benchmark for future research and provide an evidence base for policy aimed at ensuring equitable access to trustworthy, AI-assisted fact-checking.

In recent machine learning systems, confidence scores are being utilized more and more to manage selective prediction, whereby a model can abstain from making a prediction when it is unconfident. Yet, conventional metrics like accuracy, expected calibration error (ECE), and area under the risk-coverage curve (AURC) do not capture the actual reliability of predictions. These metrics either disregard confidence entirely, dilute valuable localized information through averaging, or neglect to suitably penalize overconfident misclassifications, which can be particularly detrimental in real-world systems. We introduce two new metrics Confidence-Weighted Selective Accuracy (CWSA) and its normalized variant CWSA+ that offer a principled and interpretable way to evaluate predictive models under confidence thresholds. Unlike existing methods, our metrics explicitly reward confident accuracy and penalize overconfident mistakes. They are threshold-local, decomposable, and usable in both evaluation and deployment settings where trust and risk must be quantified. Through exhaustive experiments on both real-world data sets (MNIST, CIFAR-10) and artificial model variants (calibrated, overconfident, underconfident, random, perfect), we show that CWSA and CWSA+ both effectively detect nuanced failure modes and outperform classical metrics in trust-sensitive tests. Our results confirm that CWSA is a sound basis for developing and assessing selective prediction systems for safety-critical domains.

Adversarial examples have been shown to cause neural networks to fail on a wide range of vision and language tasks, but recent work has claimed that Bayesian neural networks (bnns) are inherently robust to adversarial perturbations. In this work, we examine this claim. To study the adversarial robustness of bnns, we investigate whether it is possible to successfully break state-of-the-art bnn inference methods and prediction pipelines using even relatively unsophisticated attacks for three tasks: (1) label prediction under the posterior predictive mean, (2) adversarial example detection with Bayesian predictive uncertainty, and (3) semantic shift detection. We find that bnns trained with state-of-the-art approximate inference methods, and even bnns trained with Hamiltonian Monte Carlo, are highly susceptible to adversarial attacks. We also identify various conceptual and experimental errors in previous works that claimed inherent adversarial robustness of bnns and conclusively demonstrate that bnns and uncertainty-aware Bayesian prediction pipelines are not inherently robust against adversarial attacks.