arXiv subjects

Rodrigo C. Barros

Publications and source records attributed to Rodrigo C. Barros.

At least 19 recordsLinked to original sources

arXiv2026

Latent Fact-Checking: Detecting Misinformation through Activation Engineering

The proliferation of misinformation online has driven demand for scalable detection systems. While most existing approaches rely on surface-level linguistic features or external knowledge retrieval, we examine truthfulness as a geometric property of a language model's representation space. We introduce a misinformation detection framework grounded in activation engineering, which leverages the latent geometry of transformer models. Our approach elicits a misinformation direction in the residual stream by contrasting activations from paired truthful and false statements, following the difference-in-means principle of Contrastive Activation Addition (CAA). At inference time, the last-token activation of an unseen claim is projected onto this direction, and the projected representation is fed to an Multilayer Perceptron (MLP) for classification. The procedure requires no fine-tuning of the backbone model, no external evidence retrieval, and no task-specific supervision beyond the contrastive pairs used to estimate the direction. We evaluate the method across 11 models from the Gemma, Llama, and Qwen families, ranging from 270M to 12B parameters, on three fact-checking benchmarks: AVeriTeC, LIAR, and FACTors. The falsehood direction is recoverable across model scales and architectural families, and last-token projection matches or surpasses zero-shot and few-shot prompting baselines on LIAR and FACTors, with the largest gains observed for smaller models. Performance on AVeriTeC is more limited, which we attribute to its evidence-grounded labeling scheme. These findings provide evidence that truthfulness is a structured, linearly separable concept in the latent space of pretrained language models, and point toward interpretability-driven misinformation detection as a practical complement to retrieval-based pipelines. The code is available on https://github.com/Malta-Lab/LaFaCt.

cs.LG↗

arXiv2026

Continual Learning for Sequential Personalization of Small Language Models: A Stability Monitoring Analysis

Small Language Models (SLMs) are increasingly being considered for deployment on edge devices such as laptops, enabling private, low-latency, and locally personalized applications. However, personalization requires models to adapt over time to evolving user- or task-specific data, placing them in a continual learning setting. This creates the risk of catastrophic forgetting, where learning new information degrades performance on previously learned tasks or broader model capabilities. Recent benchmarks such as TRACE have shown that continual fine-tuning can significantly degrade the general abilities of aligned large language models. In this work, we present a study for sequential LoRA personalization of SLMs. We save model checkpoints after each adaptation stage and evaluate them on current tasks, previously seen tasks, and a fixed reference set. This checkpoint-level protocol enables us to monitor task performance, forgetting, and reference set drift over time. We show that lightweight reference set distributional diagnostics can reveal model-specific instability patterns during sequential LoRA personalization of SLMs, including cases where task-level metrics alone hide harmful adaptation. We hope this can highlight new research avenues for monitoring stability of SLMs in a continual learning setting.

cs.LG↗

arXiv2026

Low-Rank Adapters Initialization via Gradient Surgery for Continual Learning

LoRA is widely adopted for continual fine-tuning of Large Language Models due to its parameter efficiency, modularity across tasks, and compatibility with replay strategies. However, LoRA-based continual learning remains vulnerable to catastrophic forgetting, whose severity depends on how successive task gradients interact: when consecutive task gradients conflict, standard adapter initializations channel updates into subspaces that overwrite previously learned directions. We propose SLICE, a gradient-surgery-based initialization for LoRA adapters in continual learning. SLICE accumulates gradients from both the current task and a replay buffer of prior tasks, reconciles them through a projection operator, and decomposes the result via truncated SVD to initialize the adapter weights. We evaluate SLICE on the TRACE benchmark and sequences of Super-NI tasks, including a set of adversarial Super-NI sequences that we construct by mining task pairs with maximally opposing gradients. Compared to vanilla LoRA, LoRA-GA, and LoRAM, SLICE consistently achieves a better stability-plasticity trade-off, improving Average Performance, Final Performance and Forgetting metrics while preserving General Performance and In Context Performance across both standard and adversarial continual learning sequences.

cs.LG↗

arXiv2026

Inference-Time Machine Unlearning via Gated Activation Redirection

Large Language Models memorize vast amounts of training data, raising concerns regarding privacy, copyright infringement, and safety. Machine unlearning seeks to remove the influence of a targeted forget set while preserving model performance, ideally approximating a model retrained from scratch without the forget set. Existing approaches aim to achieve this by updating model parameters via gradient-based methods. However, these updates are computationally expensive, lead to irreversible weight changes, and degrade when the model is quantized for deployment. A recent alternative to changing model weights is activation engineering, where activations are changed during inference to steer model behavior. Despite circumventing weight editing, naive activation steering introduces its own failure modes, as a single global steering vector applies the same intervention to every input, leading to unintended changes in model behavior. We introduce Inference-Time Unlearning via Gated Activation Redirection (GUARD-IT), a training- and gradient-free method that unlearns via input-dependent activation steering at inference time. The resulting intervention is applied as a norm-preserving rotation in the residual stream, leaving model weights untouched. Experiments on TOFU and MUSE show that GUARD-IT matches or exceeds 12 gradient-based baselines across three model scales, while being the only method to simultaneously preserve utility, suppress memorization, and avoid catastrophic collapse across all settings. GUARD-IT further supports continual unlearning without retraining, and remains effective under quantization, a scenario in which parameter-editing methods degrade.

cs.LG↗

arXiv2026

Quantization-Robust LLM Unlearning via Low-Rank Adaptation

Large Language Model (LLM) unlearning aims to remove targeted knowledge from a trained model, but practical deployments often require post-training quantization (PTQ) for efficient inference. However, aggressive low-bit PTQ can mask unlearning updates, causing quantized models to revert to pre-unlearning behavior. We show that standard full-parameter fine-tuning often induces parameter changes that are too small to survive 4-bit quantization. We propose quantization-robust unlearning via low-rank adaptation (LoRA): we freeze the base model and concentrate unlearning into trainable adapters so that the effective update is preserved after quantization. On Llama-2-7B evaluated with MUSE dataset (BOOKS and NEWS), LoRA improves 4-bit utility by up to 7.93 points (NPO+GDR on BOOKS: 50.17 to 58.10) and yields higher 4-bit utility on NEWS for GA+GDR (40.06 to 44.82, increase of 4.76). LoRA also substantially reduces privacy leakage under 4-bit PTQ, e.g., for GA+KLR on BOOKS, PrivLeak moves from -25.68 to -5.86 (closer to ideal 0), while maintaining strong forgetting (VerMem and KnowMem near 0). Thus, using LoRA for Machine Unlearning is beneficial for scenarios where quantization is necessary for model deployment.

cs.LG↗

arXiv2025

A Framework for Benchmarking Fairness-Utility Trade-offs in Text-to-Image Models via Pareto Frontiers

Achieving fairness in text-to-image generation demands mitigating social biases without compromising visual fidelity, a challenge critical to responsible AI. Current fairness evaluation procedures for text-to-image models rely on qualitative judgment or narrow comparisons, which limit the capacity to assess both fairness and utility in these models and prevent reproducible assessment of debiasing methods. Existing approaches typically employ ad-hoc, human-centered visual inspections that are both error-prone and difficult to replicate. We propose a method for evaluating fairness and utility in text-to-image models using Pareto-optimal frontiers across hyperparametrization of debiasing methods. Our method allows for comparison between distinct text-to-image models, outlining all configurations that optimize fairness for a given utility and vice-versa. To illustrate our evaluation method, we use Normalized Shannon Entropy and ClipScore for fairness and utility evaluation, respectively. We assess fairness and utility in Stable Diffusion, Fair Diffusion, SDXL, DeCoDi, and FLUX text-to-image models. Our method shows that most default hyperparameterizations of the text-to-image model are dominated solutions in the fairness-utility space, and it is straightforward to find better hyperparameters.

cs.CV↗

arXiv2025

Inference Time Debiasing Concepts in Diffusion Models

We propose DeCoDi, a debiasing procedure for text-to-image diffusion-based models that changes the inference procedure, does not significantly change image quality, has negligible compute overhead, and can be applied in any diffusion-based image generation model. DeCoDi changes the diffusion process to avoid latent dimension regions of biased concepts. While most deep learning debiasing methods require complex or compute-intensive interventions, our method is designed to change only the inference procedure. Therefore, it is more accessible to a wide range of practitioners. We show the effectiveness of the method by debiasing for gender, ethnicity, and age for the concepts of nurse, firefighter, and CEO. Two distinct human evaluators manually inspect 1,200 generated images. Their evaluation results provide evidence that our method is effective in mitigating biases based on gender, ethnicity, and age. We also show that an automatic bias evaluation performed by the GPT4o is not significantly statistically distinct from a human evaluation. Our evaluation shows promising results, with reliable levels of agreement between evaluators and more coverage of protected attributes. Our method has the potential to significantly improve the diversity of images it generates by diffusion-based text-to-image generative models.

cs.GR↗

arXiv2025

Bayesian Attention Mechanism: A Probabilistic Framework for Positional Encoding and Context Length Extrapolation

Transformer-based language models rely on positional encoding (PE) to handle token order and support context length extrapolation. However, existing PE methods lack theoretical clarity and rely on limited evaluation metrics to substantiate their extrapolation claims. We propose the Bayesian Attention Mechanism (BAM), a theoretical framework that formulates positional encoding as a prior within a probabilistic model. BAM unifies existing methods (e.g., NoPE and ALiBi) and motivates a new Generalized Gaussian positional prior that substantially improves long-context generalization. Empirically, BAM enables accurate information retrieval at $500\times$ the training context length, outperforming previous state-of-the-art context length generalization in long context retrieval accuracy while maintaining comparable perplexity and introducing minimal additional parameters.

cs.CL↗

arXiv2023

Self-Supervised Adversarial Imitation Learning

Behavioural cloning is an imitation learning technique that teaches an agent how to behave via expert demonstrations. Recent approaches use self-supervision of fully-observable unlabelled snapshots of the states to decode state pairs into actions. However, the iterative learning scheme employed by these techniques is prone to get trapped into bad local minima. Previous work uses goal-aware strategies to solve this issue. However, this requires manual intervention to verify whether an agent has reached its goal. We address this limitation by incorporating a discriminator into the original framework, offering two key advantages and directly solving a learning problem previous work had. First, it disposes of the manual intervention requirement. Second, it helps in learning by guiding function approximation based on the state transition of the expert's trajectories. Third, the discriminator solves a learning issue commonly present in the policy model, which is to sometimes perform a `no action' within the environment until the agent finally halts.

cs.LG↗

arXiv2022

Debiasing Methods for Fairer Neural Models in Vision and Language Research: A Survey

Despite being responsible for state-of-the-art results in several computer vision and natural language processing tasks, neural networks have faced harsh criticism due to some of their current shortcomings. One of them is that neural networks are correlation machines prone to model biases within the data instead of focusing on actual useful causal relationships. This problem is particularly serious in application domains affected by aspects such as race, gender, and age. To prevent models from incurring on unfair decision-making, the AI community has concentrated efforts in correcting algorithmic biases, giving rise to the research area now widely known as fairness in AI. In this survey paper, we provide an in-depth overview of the main debiasing methods for fairness-aware neural networks in the context of vision and language research. We propose a novel taxonomy to better organize the literature on debiasing methods for fairness, and we discuss the current challenges, trends, and important future work directions for the interested researcher and practitioner.

cs.LG↗

arXiv2020

CrowdEst: A Method for Estimating (and not Simulating) Crowd Evacuation Parameters in Generic Environments

Evacuation plans have been historically used as a safety measure for the construction of buildings. The existing crowd simulators require fully-modeled 3D environments and enough time to prepare and simulate scenarios, where the distribution and behavior of the crowd needs to be controlled. In addition, its population, routes or even doors and passages may change, so the 3D model and configurations have to be updated accordingly. This is a time-consuming task that commonly has to be addressed within the crowd simulators. With that in mind, we present a novel approach to estimate the resulting data of a given evacuation scenario without actually simulating it. For such, we divide the environment into smaller modular rooms with different configurations, in a divide-and-conquer fashion. Next, we train an artificial neural network to estimate all required data regarding the evacuation of a single room. After collecting the estimated data from each room, we develop a heuristic capable of aggregating per-room information so the full environment can be properly evaluated. Our method presents an average error of 5% when compared to evacuation time in a real-life environment. Our crowd estimator approach has several advantages, such as not requiring to model the 3D environment, nor learning how to use and configure a crowd simulator, which means any user can easily use it. Furthermore, the computational time to estimate evacuation data (inference time) is virtually zero, which is much better even when compared to the best-case scenario in a real-time crowd simulator.

cs.MA↗

arXiv2020

An Extensive Experimental Evaluation of Automated Machine Learning Methods for Recommending Classification Algorithms (Extended Version)

This paper presents an experimental comparison among four Automated Machine Learning (AutoML) methods for recommending the best classification algorithm for a given input dataset. Three of these methods are based on Evolutionary Algorithms (EAs), and the other is Auto-WEKA, a well-known AutoML method based on the Combined Algorithm Selection and Hyper-parameter optimisation (CASH) approach. The EA-based methods build classification algorithms from a single machine learning paradigm: either decision-tree induction, rule induction, or Bayesian network classification. Auto-WEKA combines algorithm selection and hyper-parameter optimisation to recommend classification algorithms from multiple paradigms. We performed controlled experiments where these four AutoML methods were given the same runtime limit for different values of this limit. In general, the difference in predictive accuracy of the three best AutoML methods was not statistically significant. However, the EA evolving decision-tree induction algorithms has the advantage of producing algorithms that generate interpretable classification models and that are more scalable to large datasets, by comparison with many algorithms from other learning paradigms that can be recommended by Auto-WEKA. We also observed that Auto-WEKA has shown meta-overfitting, a form of overfitting at the meta-learning level, rather than at the base-learning level.

cs.LG↗

arXiv2020

Imitating Unknown Policies via Exploration

Behavioral cloning is an imitation learning technique that teaches an agent how to behave through expert demonstrations. Recent approaches use self-supervision of fully-observable unlabeled snapshots of the states to decode state-pairs into actions. However, the iterative learning scheme from these techniques are prone to getting stuck into bad local minima. We address these limitations incorporating a two-phase model into the original framework, which learns from unlabeled observations via exploration, substantially improving traditional behavioral cloning by exploiting (i) a sampling mechanism to prevent bad local minima, (ii) a sampling mechanism to improve exploration, and (iii) self-attention modules to capture global features. The resulting technique outperforms the previous state-of-the-art in four different environments by a large margin.

cs.LG↗

arXiv2020

HAPRec: Hybrid Activity and Plan Recognizer

Computer-based assistants have recently attracted much interest due to its applicability to ambient assisted living. Such assistants have to detect and recognize the high-level activities and goals performed by the assisted human beings. In this work, we demonstrate activity recognition in an indoor environment in order to identify the goal towards which the subject of the video is pursuing. Our hybrid approach combines an action recognition module and a goal recognition algorithm to identify the ultimate goal of the subject in the video.

cs.AI↗

arXiv2020

Component Analysis for Visual Question Answering Architectures

Recent research advances in Computer Vision and Natural Language Processing have introduced novel tasks that are paving the way for solving AI-complete problems. One of those tasks is called Visual Question Answering (VQA). A VQA system must take an image and a free-form, open-ended natural language question about the image, and produce a natural language answer as the output. Such a task has drawn great attention from the scientific community, which generated a plethora of approaches that aim to improve the VQA predictive accuracy. Most of them comprise three major components: (i) independent representation learning of images and questions; (ii) feature fusion so the model can use information from both sources to answer visual questions; and (iii) the generation of the correct answer in natural language. With so many approaches being recently introduced, it became unclear the real contribution of each component for the ultimate performance of the model. The main goal of this paper is to provide a comprehensive analysis regarding the impact of each component in VQA models. Our extensive set of experiments cover both visual and textual elements, as well as the combination of these representations in form of fusion and attention mechanisms. Our major contribution is to identify core components for training VQA models so as to maximize their predictive performance.

cs.CV↗

arXiv2019

Can we trust deep learning models diagnosis? The impact of domain shift in chest radiograph classification

While deep learning models become more widespread, their ability to handle unseen data and generalize for any scenario is yet to be challenged. In medical imaging, there is a high heterogeneity of distributions among images based on the equipment that generates them and their parametrization. This heterogeneity triggers a common issue in machine learning called domain shift, which represents the difference between the training data distribution and the distribution of where a model is employed. A high domain shift tends to implicate in a poor generalization performance from the models. In this work, we evaluate the extent of domain shift on four of the largest datasets of chest radiographs. We show how training and testing with different datasets (e.g., training in ChestX-ray14 and testing in CheXpert) drastically affects model performance, posing a big question over the reliability of deep learning models trained on public datasets. We also show that models trained on CheXpert and MIMIC-CXR generalize better to other datasets.

eess.IV↗

arXiv2019

Classifying Norm Conflicts using Learned Semantic Representations

While most social norms are informal, they are often formalized by companies in contracts to regulate trades of goods and services. When poorly written, contracts may contain normative conflicts resulting from opposing deontic meanings or contradict specifications. As contracts tend to be long and contain many norms, manually identifying such conflicts requires human-effort, which is time-consuming and error-prone. Automating such task benefits contract makers increasing productivity and making conflict identification more reliable. To address this problem, we introduce an approach to detect and classify norm conflicts in contracts by converting them into latent representations that preserve both syntactic and semantic information and training a model to classify norm conflicts in four conflict types. Our results reach the new state of the art when compared to a previous approach.

cs.CL↗

arXiv2019

Unsupervised domain adaptation for medical imaging segmentation with self-ensembling

Recent advances in deep learning methods have come to define the state-of-the-art for many medical imaging applications, surpassing even human judgment in several tasks. Those models, however, when trained to reduce the empirical risk on a single domain, fail to generalize when applied to other domains, a very common scenario in medical imaging due to the variability of images and anatomical structures, even across the same imaging modality. In this work, we extend the method of unsupervised domain adaptation using self-ensembling for the semantic segmentation task and explore multiple facets of the method on a small and realistic publicly-available magnetic resonance (MRI) dataset. Through an extensive evaluation, we show that self-ensembling can indeed improve the generalization of the models even when using a small amount of unlabelled data.

cs.CV↗