arXiv subjects

Joao Fonseca

Publications and source records attributed to Joao Fonseca.

11 recordsLinked to original sources

arXiv2026

Mixture-of-Expert Blocks Contain Strong Hallucination Detection Signals

Despite their widespread use, Large Language Models (LLMs) remain limited by a fundamental problem: the generation of plausible but false content, known as hallucinations. Most existing detection methods operate at the answer or sentence level, yet per-token detection is essential for localizing hallucinated spans and enabling fine-grained interventions. In this paper, we explore the use of the Mixture-of-Experts (MoE) paradigm to address this gap. In MoE architectures, a single forward pass activates a sparse subset of experts (i.e., distinct feedforward networks per layer) via a routing mechanism, producing internal signals (e.g., router entropy, expert disagreement, and expert usage patterns) that are unavailable in dense architectures and have not been previously exploited for hallucination detection. To this end, we introduce InnerExpert, the first method to leverage these MoE-specific signals for per-token hallucination detection. InnerExpert combines routing-level and standard transformer signals into compact per-token feature vectors, classified by a lightweight detector trained on labels produced by an LLM-as-a-judge pipeline, which enables continuous model updates without manual annotation. Our results show that InnerExpert outperforms existing methods across five datasets and two MoE architectures, achieving up to 0.91 answer-level and 0.76 token-level AUROC, while requiring only a single forward pass.

cs.AI↗

arXiv2026

RelShap: Relationally Consistent Shapley Explanations

Machine learning pipelines commonly flatten relational data into single-table representations, discarding structural constraints. Widely used Shapley value-based feature attributions then rely on feature independence, evaluating the model on combinations that could never arise in the underlying data, producing misleading explanations. We propose RelShap, a framework that incorporates relational constraints and data provenance into Shapley value computation, restricting both background data and coalition evaluation to relationally valid configurations. The framework is estimator-agnostic and composes with Kernel SHAP, Monte Carlo, and Leverage SHAP without altering their sampling or weighting properties. Functional dependencies further induce equivalence classes over feature coalitions, which RelShap exploits to reduce runtime without changing Shapley values; we provide a combinatorial characterization of the expected speedup. Experiments across multiple datasets, models, and estimators show that RelShap produces explanations that are more faithful to the data-generating process, correctly identifying the dominant feature in controlled settings where existing methods, including Conditional SHAP and ManifoldShap, do not. Our code is available at: https://github.com/duneag2/relshap.

cs.LG↗

arXiv2026

ExplainerPFN: Towards tabular foundation models for model-free zero-shot feature importance estimations

Computing the importance of features in supervised classification tasks is critical for model interpretability. Shapley values are a widely used approach for explaining model predictions, but require direct access to the underlying model, an assumption frequently violated in real-world deployments. We investigate whether meaningful feature attributions can be obtained in a zero-shot setting, using only the input data distribution and no evaluations of the target model. Because multiple models can produce identical predictions yet yield different Shapley decompositions, the mapping from data to attributions is not uniquely identifiable. We therefore target attributions that are "true to the data" rather than "true to the model", learning a posterior mean attribution under a meta-training prior. To this end, we introduce ExplainerPFN, a tabular foundation model built on TabPFN, pretrained on synthetic structural causal datasets supervised with exact or near-exact Shapley values, that predicts feature attributions for unseen tabular datasets without model access, gradients, or example explanations. Our contributions are fourfold: (1) we show that few-shot surrogate explainers achieve high SHAP fidelity with as few as two reference observations; (2) we propose ExplainerPFN, the first zero-shot method for estimating Shapley-value-style feature attributions without access to the underlying model or reference explanations, providing a principled attribution where no existing explainer can be applied; (3) we release an open-source implementation including the full training pipeline and synthetic data generator; and (4) through extensive experiments on real and synthetic datasets, we show that ExplainerPFN achieves performance competitive with few-shot surrogate explainers that rely on 2-10 SHAP examples.

cs.LG↗

arXiv2025

ShaRP: Explaining Rankings and Preferences with Shapley Values

Algorithmic decisions in critical domains such as hiring, college admissions, and lending are often based on rankings. Given the impact of these decisions on individuals, organizations, and population groups, it is essential to understand them - to help individuals improve their ranking position, design better ranking procedures, and ensure legal compliance. In this paper, we argue that explainability methods for classification and regression, such as SHAP, are insufficient for ranking tasks, and present ShaRP - Shapley Values for Rankings and Preferences - a framework that explains the contributions of features to various aspects of a ranked outcome. ShaRP computes feature contributions for various ranking-specific profit functions, such as rank and top-k, and also includes a novel Shapley value-based method for explaining pairwise preference outcomes. We provide a flexible implementation of ShaRP, capable of efficiently and comprehensively explaining ranked and pairwise outcomes over tabular data, in score-based ranking and learning-to-rank tasks. Finally, we develop a comprehensive evaluation methodology for ranking explainability methods, showing through qualitative, quantitative, and usability studies that our rank-aware QoIs offer complementary insights, scale effectively, and help users interpret ranked outcomes in practice.

cs.AI↗

arXiv2025

SHAP-based Explanations are Sensitive to Feature Representation

Local feature-based explanations are a key component of the XAI toolkit. These explanations compute feature importance values relative to an ``interpretable'' feature representation. In tabular data, feature values themselves are often considered interpretable. This paper examines the impact of data engineering choices on local feature-based explanations. We demonstrate that simple, common data engineering techniques, such as representing age with a histogram or encoding race in a specific way, can manipulate feature importance as determined by popular methods like SHAP. Notably, the sensitivity of explanations to feature representation can be exploited by adversaries to obscure issues like discrimination. While the intuition behind these results is straightforward, their systematic exploration has been lacking. Previous work has focused on adversarial attacks on feature-based explainers by biasing data or manipulating models. To the best of our knowledge, this is the first study demonstrating that explainers can be misled by standard, seemingly innocuous data engineering techniques.

cs.LG↗

arXiv2025

Output Scouting: Auditing Large Language Models for Catastrophic Responses

Recent high profile incidents in which the use of Large Language Models (LLMs) resulted in significant harm to individuals have brought about a growing interest in AI safety. One reason LLM safety issues occur is that models often have at least some non-zero probability of producing harmful outputs. In this work, we explore the following scenario: imagine an AI safety auditor is searching for catastrophic responses from an LLM (e.g. a "yes" responses to "can I fire an employee for being pregnant?"), and is able to query the model a limited number times (e.g. 1000 times). What is a strategy for querying the model that would efficiently find those failure responses? To this end, we propose output scouting: an approach that aims to generate semantically fluent outputs to a given prompt matching any target probability distribution. We then run experiments using two LLMs and find numerous examples of catastrophic responses. We conclude with a discussion that includes advice for practitioners who are looking to implement LLM auditing for catastrophic responses. We also release an open-source toolkit (https://github.com/joaopfonseca/outputscouting) that implements our auditing framework using the Hugging Face transformers library.

cs.CL↗

arXiv2025

Safeguarding Large Language Models in Real-time with Tunable Safety-Performance Trade-offs

Large Language Models (LLMs) have been shown to be susceptible to jailbreak attacks, or adversarial attacks used to illicit high risk behavior from a model. Jailbreaks have been exploited by cybercriminals and blackhat actors to cause significant harm, highlighting the critical need to safeguard widely-deployed models. Safeguarding approaches, which include fine-tuning models or having LLMs "self-reflect", may lengthen the inference time of a model, incur a computational penalty, reduce the semantic fluency of an output, and restrict ``normal'' model behavior. Importantly, these Safety-Performance Trade-offs (SPTs) remain an understudied area. In this work, we introduce a novel safeguard, called SafeNudge, that combines Controlled Text Generation with "nudging", or using text interventions to change the behavior of a model. SafeNudge triggers during text-generation while a jailbreak attack is being executed, and can reduce successful jailbreak attempts by 30% by guiding the LLM towards a safe responses. It adds minimal latency to inference and has a negligible impact on the semantic fluency of outputs. Further, we allow for tunable SPTs. SafeNudge is open-source and available through https://pypi.org/, and is compatible with models loaded with the Hugging Face "transformers" library.

cs.CL↗

arXiv2024

Development of Machine Learning Classifiers for Blood-based Diagnosis and Prognosis of Suspected Acute Infections and Sepsis

We applied machine learning to the unmet medical need of rapid and accurate diagnosis and prognosis of acute infections and sepsis in emergency departments. Our solution consists of a Myrna (TM) Instrument and embedded TriVerity (TM) classifiers. The instrument measures abundances of 29 messenger RNAs in patient's blood, subsequently used as features for machine learning. The classifiers convert the input features to an intuitive test report comprising the separate likelihoods of (1) a bacterial infection (2) a viral infection, and (3) severity (need for Intensive Care Unit-level care). In internal validation, the system achieved AUROC = 0.83 on the three-class disease diagnosis (bacterial, viral, or non-infected) and AUROC = 0.77 on binary prognosis of disease severity. The Myrna, TriVerity system was granted breakthrough device designation by the United States Food and Drug Administration (FDA). This engineering manuscript teaches the standard and novel machine learning methods used to translate an academic research concept to a clinical product aimed at improving patient care, and discusses lessons learned.

q-bio.QM↗

arXiv2024

Fairness in Algorithmic Recourse Through the Lens of Substantive Equality of Opportunity

Algorithmic recourse -- providing recommendations to those affected negatively by the outcome of an algorithmic system on how they can take action and change that outcome -- has gained attention as a means of giving persons agency in their interactions with artificial intelligence (AI) systems. Recent work has shown that even if an AI decision-making classifier is ``fair'' (according to some reasonable criteria), recourse itself may be unfair due to differences in the initial circumstances of individuals, compounding disparities for marginalized populations and requiring them to exert more effort than others. There is a need to define more methods and metrics for evaluating fairness in recourse that span a range of normative views of the world, and specifically those that take into account time. Time is a critical element in recourse because the longer it takes an individual to act, the more the setting may change due to model or data drift. This paper seeks to close this research gap by proposing two notions of fairness in recourse that are in normative alignment with substantive equality of opportunity, and that consider time. The first considers the (often repeated) effort individuals exert per successful recourse event, and the second considers time per successful recourse event. Building upon an agent-based framework for simulating recourse, this paper demonstrates how much effort is needed to overcome disparities in initial circumstances. We then proposes an intervention to improve the fairness of recourse by rewarding effort, and compare it to existing strategies.

cs.LG↗

arXiv2023

Setting the Right Expectations: Algorithmic Recourse Over Time

Algorithmic systems are often called upon to assist in high-stakes decision making. In light of this, algorithmic recourse, the principle wherein individuals should be able to take action against an undesirable outcome made by an algorithmic system, is receiving growing attention. The bulk of the literature on algorithmic recourse to-date focuses primarily on how to provide recourse to a single individual, overlooking a critical element: the effects of a continuously changing context. Disregarding these effects on recourse is a significant oversight, since, in almost all cases, recourse consists of an individual making a first, unfavorable attempt, and then being given an opportunity to make one or several attempts at a later date - when the context might have changed. This can create false expectations, as initial recourse recommendations may become less reliable over time due to model drift and competition for access to the favorable outcome between individuals. In this work we propose an agent-based simulation framework for studying the effects of a continuously changing environment on algorithmic recourse. In particular, we identify two main effects that can alter the reliability of recourse for individuals represented by the agents: (1) competition with other agents acting upon recourse, and (2) competition with new agents entering the environment. Our findings highlight that only a small set of specific parameterizations result in algorithmic recourse that is reliable for agents over time. Consequently, we argue that substantial additional work is needed to understand recourse reliability over time, and to develop recourse methods that reward agents' effort.

cs.LG↗

arXiv2022

Research Trends and Applications of Data Augmentation Algorithms

In the Machine Learning research community, there is a consensus regarding the relationship between model complexity and the required amount of data and computation power. In real world applications, these computational requirements are not always available, motivating research on regularization methods. In addition, current and past research have shown that simpler classification algorithms can reach state-of-the-art performance on computer vision tasks given a robust method to artificially augment the training dataset. Because of this, data augmentation techniques became a popular research topic in recent years. However, existing data augmentation methods are generally less transferable than other regularization methods. In this paper we identify the main areas of application of data augmentation algorithms, the types of algorithms used, significant research trends, their progression over time and research gaps in data augmentation literature. To do this, the related literature was collected through the Scopus database. Its analysis was done following network science, text mining and exploratory analysis approaches. We expect readers to understand the potential of data augmentation, as well as identify future research directions and open questions within data augmentation research.

cs.LG↗