arXiv subjects

Owen O'Neill

Publications and source records attributed to Owen O'Neill.

4 recordsLinked to original sources

arXiv2026

From Tokens to Semantics: Leveraging Complementary Signals for Hallucination Detection in Black-Box LLMs

When LLMs support public-facing or high-stakes workflows, missed fabrications can harm users and institutions, while false alarms consume limited human-review capacity. When no trusted context or reference document is available, we study two signals accessible through black-box model APIs: semantic entropy, which measures disagreement among sampled response meanings, and uncertainty derived from token log-probabilities. Their failure modes can be complementary: semantic entropy becomes uninformative when responses form one semantic cluster, while token uncertainty can miss consistently confident errors. We extend token-based uncertainty detection by aggregating token-level signals across sampled responses through our TopK method, evaluate the hybrid CoCoA method, which combines target-response uncertainty with semantic dissimilarity, and propose and study two supervised methods: Gated, which routes single-cluster cases to an aggregated-token-feature classifier, and Stacked, which learns jointly from semantic uncertainty and broader token features. We evaluate seven benchmarks, including five public benchmarks (four text datasets and multimodal handwritten-cheque extraction) and two constructed benchmarks (Financial Summaries and Long-Text QA), using four language models. In our evaluation across models and datasets, Stacked gave the best performance in nearly half of the cases, while TopK and CoCoA remain competitive without supervised training labels, although their thresholds require careful calibration. No method is universally strongest. We therefore evaluate performance at false-positive-rate budgets from 1% to 15%, assess their sensitivity to generation and calibration choices, and examine variation across dataset characteristics.

cs.CL↗

arXiv2026

Necessary or Sufficient? Evaluating LLM Explanations With Behavioural Evidence

LLM decision components that can operate within agent workflows often produce action-relevant recommendations or judgements together with explanations. Operators may use the named factors to monitor a system, diagnose errors, or decide when to escalate an output. Such use assumes that the explanations agree with the component's observable decision behaviour. We test two interpretations of the named factors: necessity, meaning that changing a factor would change the output, and sufficiency, meaning that retaining it while removing other changeable information would preserve the output. We evaluate these interpretations in two synthetic use cases: recommending advisors to clients and judging prompts for harmfulness or risk. Models return an output and the top three factors that most influenced it. Controlled black-box interventions estimate a necessity score for each factor by measuring how often changing it changes the output, and a sufficiency score by measuring how often retaining it preserves the output. Across eight models from the Claude, GPT, and Gemini families, the mean Spearman correlations between the cited ranking and the necessity and sufficiency scores are 0.349 and 0.354 for advisor recommendation, and 0.431 and 0.580 for prompt monitoring. Furthermore, an uncited factor scores above the lowest-scoring cited factor in 57.6% of advisor responses under necessity and 58.1% under sufficiency; the corresponding prompt-monitoring rates are 25.8% and 8.9%. The cited top three contain useful information but do not reliably identify the three factors with the strongest measured influence under necessity or sufficiency. The framework provides a black-box reliability check for explanations used in agent oversight while remaining scoped to individual LLM decisions.

cs.AI↗

arXiv2026

Towards Provably Fair Machine Learning: Bayesian Approaches For Consistent and Transparent Predictions

ML classifiers deployed in high-stakes domains produce predictions whose quality varies systematically across subgroups. For granular subgroups defined by intersections of multiple features, predictions are often inconsistent with the observed data: the model's outputs contradict the evidence available for that subgroup. This problem is exacerbated by regularisation, which improves aggregate performance by collapsing small subgroups into larger groups, disproportionately affecting demographic minorities. We define two requirements for consistent prediction: determinism (identical individuals receive identical predictions) and statistical consistency (we cannot reject, at significance level alpha, the hypothesis that the predictions for a subgroup were drawn from the Bayesian optimal target distribution inferred for that subgroup). From these requirements we derive the Fair Bayesian classifier, which enforces both across every group and subgroup simultaneously and abstains whenever no consistent deterministic prediction is possible. On three benchmark datasets (Adult, COMPAS, and Bank Marketing), standard classifiers produce statistically inconsistent predictions for a substantial proportion of subgroups. Our classifier achieves zero consistency error by construction while exceeding baseline accuracy and multicalibration on every dataset tested. Statistical consistency provides a principled foundation for prediction quality with direct implications for algorithmic fairness. Minority demographics are disproportionately concentrated in small subgroups, precisely where frequentist inference is least reliable; addressing this inference problem is therefore a necessary step toward fair ML. By enforcing Bayesian consistency at the finest resolution the data supports, the our classifier demonstrates that exhaustive subgroup fairness with principled abstention is achievable in practice.

cs.LG↗

arXiv2023

Systematic Bias in Sample Inference and its Effect on Machine Learning

A commonly observed pattern in machine learning models is an underprediction of the target feature, with the model's predicted target rate for members of a given category typically being lower than the actual target rate for members of that category in the training set. This underprediction is usually larger for members of minority groups; while income level is underpredicted for both men and women in the 'adult' dataset, for example, the degree of underprediction is significantly higher for women (a minority in that dataset). We propose that this pattern of underprediction for minorities arises as a predictable consequence of statistical inference on small samples. When presented with a new individual for classification, an ML model performs inference not on the entire training set, but on a subset that is in some way similar to the new individual, with sizes of these subsets typically following a power law distribution so that most are small (and with these subsets being necessarily smaller for the minority group). We show that such inference on small samples is subject to systematic and directional statistical bias, and that this bias produces the observed patterns of underprediction seen in ML models. Analysing a standard sklearn decision tree model's predictions on a set of over 70 subsets of the 'adult' and COMPAS datasets, we found that a bias prediction measure based on small-sample inference had a significant positive correlations (0.56 and 0.85) with the observed underprediction rate for these subsets.

cs.LG↗