arXiv subjects

Lucas Craig

Publications and source records attributed to Lucas Craig.

3 recordsLinked to original sources

arXiv2026

Beyond Confidence: Rethinking Self-Assessments for Performance Prediction in LLMs

Large Language Models (LLMs) are increasingly used in settings where reliable self-assessment is critical. Assessing model reliability has evolved from using probabilistic correctness estimates to, more recently, eliciting verbalized confidence. Confidence, however, has been shown to be an inconsistent and overoptimistic predictor of model correctness. Drawing on cognitive appraisal theory, a framework from human psychology that decomposes self-evaluation into multiple components, we propose a multidimensional perspective on model self-assessment. We elicit six appraisal-based dimensions of self-assessment, alongside confidence, and evaluate their utility for predicting model failure across 12 LLMs and 38 tasks spanning eight domains. We find that competence-related appraisal dimensions, particularly effort and ability, consistently match or outperform confidence across most settings. Effort additionally yields less overoptimistic estimates that remain stable across model sizes. In contrast, affective dimensions provide marginally predictive signals. Furthermore, the most informative dimension varies systematically with task characteristics: effort is most predictive for reasoning-intensive tasks, while ability and confidence dominate on retrieval-oriented tasks. Broadly, our findings indicate that structured multidimensional self-assessment is a promising approach to improving the reliability and safety of language model deployment across diverse real-world settings.

cs.CL↗

arXiv2025

Large language models show fragile cognitive reasoning about human emotions

Affective computing seeks to support the holistic development of artificial intelligence by enabling machines to engage with human emotion. Recent foundation models, particularly large language models (LLMs), have been trained and evaluated on emotion-related tasks, typically using supervised learning with discrete emotion labels. Such evaluations largely focus on surface phenomena, such as recognizing expressed or evoked emotions, leaving open whether these systems reason about emotion in cognitively meaningful ways. Here we ask whether LLMs can reason about emotions through underlying cognitive dimensions rather than labels alone. Drawing on cognitive appraisal theory, we introduce CoRE, a large-scale benchmark designed to probe the implicit cognitive structures LLMs use when interpreting emotionally charged situations. We assess alignment with human appraisal patterns, internal consistency, cross-model generalization, and robustness to contextual variation. We find that LLMs capture systematic relations between cognitive appraisals and emotions but show misalignment with human judgments and instability across contexts.

cs.CL↗

arXiv2025

Disparate Privacy Vulnerability: Targeted Attribute Inference Attacks and Defenses

As machine learning (ML) technologies become more prevalent in privacy-sensitive areas like healthcare and finance, eventually incorporating sensitive information in building data-driven algorithms, it is vital to scrutinize whether these data face any privacy leakage risks. One potential threat arises from an adversary querying trained models using the public, non-sensitive attributes of entities in the training data to infer their private, sensitive attributes, a technique known as the attribute inference attack. This attack is particularly deceptive because, while it may perform poorly in predicting sensitive attributes across the entire dataset, it excels at predicting the sensitive attributes of records from a few vulnerable groups, a phenomenon known as disparate vulnerability. This paper illustrates that an adversary can take advantage of this disparity to carry out a series of new attacks, showcasing a threat level beyond previous imagination. We first develop a novel inference attack called the disparity inference attack, which targets the identification of high-risk groups within the dataset. We then introduce two targeted variations of the attribute inference attack that can identify and exploit a vulnerable subset of the training data, marking the first instances of targeted attacks in this category, achieving significantly higher accuracy than untargeted versions. We are also the first to introduce a novel and effective disparity mitigation technique that simultaneously preserves model performance and prevents any risk of targeted attacks.

cs.LG↗