arXiv subjects

Alfio Ferrara

Publications and source records attributed to Alfio Ferrara.

16 recordsLinked to original sources

arXiv2026

LumiXAI: A Modular Full-Stack Framework for Feature Attribution

Feature attribution is a central tool of model interpretability, yet the software through which it is applied remains fragmented: individual tools specialize along narrow axes, such as a single modality, a code API or a GUI, or a fixed rather than extensible method set, and rarely combine these strengths. Moreover, many explainability tools are designed primarily for domain experts, requiring programming skills or familiarity with attribution methods that can make them difficult for non-expert users to access. In this article, we present LumiXAI, a modular full-stack framework that consolidates attribution analysis into a single system. It couples classification and generative attribution with an interactive GUI supporting bidirectional exploration, a plug-in architecture for registering new models and methods, and three access tiers serving non-programmers, developers, and extenders from one backend. Its contribution is a system that operationalises established attribution methods under one interface, one interaction model, and one persistence layer, with containerised services and persistent results making analyses reproducible across machines.

cs.SE↗

arXiv2026

Refusal Beyond a Single Direction: A Preliminary Comparison of Diff-in-Means and INLP

Arditi et al. (2024) has shown that refusal in safety fine-tuned chat models is mediated by a single linear direction in the residual stream, recoverable by a difference-in-means (DiM) of harmful and harmless activations. We compare DiM-based interventions (activation addition and directional ablation) with two interventions derived from Iterative Nullspace Projection (INLP)-nullspace projection and counterfactual flipping-on five open-weight chat models, asking whether INLP can match DiM at steering refusal and whether its richer parameterisation yields more tweakable interventions. INLP counterfactual flipping is competitive with DiM directional ablation on refusal suppression, while nullspace projection is weaker on most models. Applying each method at the layer selected by the other shows that flipping's competitiveness is robust to this change while directional ablation's is not, suggesting that DiM's apparent edge is partly a property of its own selection procedure. Restricting INLP to the leading directions of the extracted subspace preserves most of the suppression effect at near-baseline perplexity, giving a tunable capability. Geometrically, the two INLP interventions land in qualitatively different regions of activation space: nullspace projection collapses transformed activations between the harmful and harmless clusters, while counterfactual flipping moves them into the opposite cluster, suggesting that the model encodes the absence of a concept differently from its opposit--an intriguing distinction that warrants further investigation in future work.

cs.AI↗

arXiv2026

How LLMs Follow Instructions: Skillful Coordination, Not a Universal Mechanism

Instruction tuning is commonly assumed to endow language models with a domain-general ability to follow instructions, yet the underlying mechanism remains poorly understood. Does instruction-following rely on a universal mechanism or compositional skill deployment? We investigate this through diagnostic probing across nine diverse tasks in three instruction-tuned models. Our analysis provides converging evidence against a universal mechanism. First, general probes trained across all tasks show selective rather than uniform deficits relative to task-specific specialists, indicating that representational sharing is partial and structured rather than global. Second, cross-task transfer is weak and clustered by skill similarity. Third, causal ablation reveals sparse asymmetric dependencies rather than shared representations. Tasks also stratify by complexity across layers, with structural constraints emerging early and semantic tasks emerging late. Finally, temporal analysis shows that the constraint signal becomes decodable only once generation is under way, and remains so throughout the response. These findings indicate that instruction-following is better characterized as skillful coordination of diverse linguistic capabilities rather than deployment of a single abstract constraint-checking process.

cs.AI↗

arXiv2026

Transcription and Recognition of Italian Parliamentary Speeches Using Vision-Language Models

Parliamentary proceedings represent a rich yet challenging resource for computational analysis, particularly when preserved only as scanned historical documents. Existing efforts to transcribe Italian parliamentary speeches have relied on traditional Optical Character Recognition pipelines, resulting in transcription errors and limited semantic annotation. In this paper, we propose a pipeline based on Vision-Language Models for the automatic transcription, semantic segmentation, and entity linking of Italian parliamentary speeches. The pipeline employs a specialised OCR model to extract text while preserving reading order, followed by a large-scale Vision-Language Model that performs transcription refinement, element classification, and speaker identification by jointly reasoning over visual layout and textual content. Extracted speakers are then linked to the Chamber of Deputies knowledge base through SPARQL queries and a multi-strategy fuzzy matching procedure. Evaluation against an established benchmark demonstrates substantial improvements both in transcription quality and speaker tagging.

cs.DL↗

arXiv2026

Quid est VERITAS? A Modular Framework for Archival Document Analysis

The digitisation of historical documents has traditionally been conceived as a process limited to character-level transcription, producing flat text that lacks the structural and semantic information necessary for substantive computational analysis. We present VERITAS (Vision-Enhanced Reading, Interpretation, and Transcription of Archival Sources), a modular, model-agnostic framework that reconceptualises digitisation as an integrated workflow encompassing transcription, layout analysis, and semantic enrichment. The pipeline is organised into four stages - Preprocessing, Extraction, Refinement, and Enrichment - and employs a schema-driven architecture that allows researchers to declaratively specify their extraction objectives. We evaluate VERITAS on the critical edition of Bernardino Corio's Storia di Milano, a Renaissance chronicle of over 1,600 pages. Results demonstrate that the pipeline achieves a 67.6% relative reduction in word error rate compared to a commercial OCR baseline, with a threefold reduction in end-to-end processing time when accounting for manual correction. We further illustrate the downstream utility of the pipeline's output by querying the transcribed corpus through a retrieval-augmented generation system, demonstrating its capacity to support historical inquiry.

cs.DL↗

arXiv2025

The Invisible Handshake: Persistent Overpricing by Adaptive Market Agents

We study overpricing in a repeated game between two representative agents: a market maker, who controls market liquidity, and a market taker, who chooses trade quantities. Market prices evolve through the endogenous price impact of trades and exogenous shocks. We define overpricing relative to a counterfactual price path that holds fixed the same sequence of shocks while shutting down price impact, and characterize the set of feasible strategy profiles that generate persistent overpricing while respecting cash and inventory constraints. We provide a sufficient condition for decentralized learning to reach the overpricing region in finite time, and we show that this condition is satisfied, in particular, by projected stochastic gradient ascent. A key step in the analysis is a decomposition of the game into a competitive component, which favors zero price impact, and a collaborative component, which makes overpricing jointly profitable when aggregate inventory is positive. We further show that the same structural incentives govern both myopic and farsighted objectives. Together, these results show how decentralized learning by adaptive market agents can lead to persistent overpricing in financial markets.

q-fin.TR↗

arXiv2025

How Instruction-Tuning Imparts Length Control: A Cross-Lingual Mechanistic Analysis

Adhering to explicit length constraints, such as generating text with a precise word count, remains a significant challenge for Large Language Models (LLMs). This study aims at investigating the differences between foundation models and their instruction-tuned counterparts, on length-controlled text generation in English and Italian. We analyze both performance and internal component contributions using Cumulative Weighted Attribution, a metric derived from Direct Logit Attribution. Our findings reveal that instruction-tuning substantially improves length control, primarily by specializing components in deeper model layers. Specifically, attention heads in later layers of IT models show increasingly positive contributions, particularly in English. In Italian, while attention contributions are more attenuated, final-layer MLPs exhibit a stronger positive role, suggesting a compensatory mechanism. These results indicate that instruction-tuning reconfigures later layers for task adherence, with component-level strategies potentially adapting to linguistic context.

cs.CL↗

arXiv2025

What's Taboo for You? - An Empirical Evaluation of LLMs Behavior Toward Sensitive Content

Proprietary Large Language Models (LLMs) have shown tendencies toward politeness, formality, and implicit content moderation. While previous research has primarily focused on explicitly training models to moderate and detoxify sensitive content, there has been limited exploration of whether LLMs implicitly sanitize language without explicit instructions. This study empirically analyzes the implicit moderation behavior of GPT-4o-mini when paraphrasing sensitive content and evaluates the extent of sensitivity shifts. Our experiments indicate that GPT-4o-mini systematically moderates content toward less sensitive classes, with substantial reductions in derogatory and taboo language. Also, we evaluate the zero-shot capabilities of LLMs in classifying sentence sensitivity, comparing their performances against traditional methods.

cs.CL↗

arXiv2025

The Cow of Rembrandt - Analyzing Artistic Prompt Interpretation in Text-to-Image Models

Text-to-image diffusion models have demonstrated remarkable capabilities in generating artistic content by learning from billions of images, including popular artworks. However, the fundamental question of how these models internally represent concepts, such as content and style in paintings, remains unexplored. Traditional computer vision assumes content and style are orthogonal, but diffusion models receive no explicit guidance about this distinction during training. In this work, we investigate how transformer-based text-to-image diffusion models encode content and style concepts when generating artworks. We leverage cross-attention heatmaps to attribute pixels in generated images to specific prompt tokens, enabling us to isolate image regions influenced by content-describing versus style-describing tokens. Our findings reveal that diffusion models demonstrate varying degrees of content-style separation depending on the specific artistic prompt and style requested. In many cases, content tokens primarily influence object-related regions while style tokens affect background and texture areas, suggesting an emergent understanding of the content-style distinction. These insights contribute to our understanding of how large-scale generative models internally represent complex artistic concepts without explicit supervision. We share the code and dataset, together with an exploratory tool for visualizing attention maps at https://github.com/umilISLab/artistic-prompt-interpretation.

cs.CV↗

arXiv2024

Unveiling Transformer Perception by Exploring Input Manifolds

This paper introduces a general method for the exploration of equivalence classes in the input space of Transformer models. The proposed approach is based on sound mathematical theory which describes the internal layers of a Transformer architecture as sequential deformations of the input manifold. Using eigendecomposition of the pullback of the distance metric defined on the output space through the Jacobian of the model, we are able to reconstruct equivalence classes in the input space and navigate across them. Our method enables two complementary exploration procedures: the first retrieves input instances that produce the same class probability distribution as the original instance-thus identifying elements within the same equivalence class-while the second discovers instances that yield a different class probability distribution, effectively navigating toward distinct equivalence classes. Finally, we demonstrate how the retrieved instances can be meaningfully interpreted by projecting their embeddings back into a human-readable format.

cs.LG↗

arXiv2024

LiMe: a Latin Corpus of Late Medieval Criminal Sentences

The Latin language has received attention from the computational linguistics research community, which has built, over the years, several valuable resources, ranging from detailed annotated corpora to sophisticated tools for linguistic analysis. With the recent advent of large language models, researchers have also started developing models capable of generating vector representations of Latin texts. The performances of such models remain behind the ones for modern languages, given the disparity in available data. In this paper, we present the LiMe dataset, a corpus of 325 documents extracted from a series of medieval manuscripts called Libri sententiarum potestatis Mediolani, and thoroughly annotated by experts, in order to be employed for masked language model, as well as supervised natural language processing tasks.

cs.CL↗

arXiv2024

How BERT Speaks Shakespearean English? Evaluating Historical Bias in Contextual Language Models

In this paper, we explore the idea of analysing the historical bias of contextual language models based on BERT by measuring their adequacy with respect to Early Modern (EME) and Modern (ME) English. In our preliminary experiments, we perform fill-in-the-blank tests with 60 masked sentences (20 EME-specific, 20 ME-specific and 20 generic) and three different models (i.e., BERT Base, MacBERTh, English HLM). We then rate the model predictions according to a 5-point bipolar scale between the two language varieties and derive a weighted score to measure the adequacy of each model to EME and ME varieties of English.

cs.CL↗

arXiv2024

Incremental Affinity Propagation based on Cluster Consolidation and Stratification

Modern data mining applications require to perform incremental clustering over dynamic datasets by tracing temporal changes over the resulting clusters. In this paper, we propose A-Posteriori affinity Propagation (APP), an incremental extension of Affinity Propagation (AP) based on cluster consolidation and cluster stratification to achieve faithfulness and forgetfulness. APP enforces incremental clustering where i) new arriving objects are dynamically consolidated into previous clusters without the need to re-execute clustering over the entire dataset of objects, and ii) a faithful sequence of clustering results is produced and maintained over time, while allowing to forget obsolete clusters with decremental learning functionalities. Four popular labeled datasets are used to test the performance of APP with respect to benchmark clustering performances obtained by conventional AP and Incremental Affinity Propagation based on Nearest neighbor Assignment (IAPNA) algorithms. Experimental results show that APP achieves comparable clustering performance while enforcing scalability at the same time.

cs.LG↗

arXiv2021

An Explainable Probabilistic Classifier for Categorical Data Inspired to Quantum Physics

This paper presents Sparse Tensor Classifier (STC), a supervised classification algorithm for categorical data inspired by the notion of superposition of states in quantum physics. By regarding an observation as a superposition of features, we introduce the concept of wave-particle duality in machine learning and propose a generalized framework that unifies the classical and the quantum probability. We show that STC possesses a wide range of desirable properties not available in most other machine learning methods but it is at the same time exceptionally easy to comprehend and use. Empirical evaluation of STC on structured data and text classification demonstrates that our methodology achieves state-of-the-art performances compared to both standard classifiers and deep learning, at the additional benefit of requiring minimal data pre-processing and hyper-parameter tuning. Moreover, STC provides a native explanation of its predictions both for single instances and for each target label globally.

cs.LG↗

arXiv2016

Research infrastructures in the LHC era: a scientometric approach

When a research infrastructure is funded and implemented, new information and new publications are created. This new information is the measurable output of discovery process. In this paper, we describe the impact of infrastructure for physics experiments in terms of publications and citations. In particular, we consider the Large Hadron Collider (LHC) experiments (ATLAS, CMS, ALICE, LHCb) and compare them to the Large Electron Positron Collider (LEP) experiments (ALEPH, DELPHI, L3, OPAL) and the Tevatron experiments (CDF, D0). We provide an overview of the scientific output of these projects over time and highlight the role played by remarkable project results in the publication-citation distribution trends. The methodological and technical contribution of this work provides a starting point for the development of a theoretical model of modern scientific knowledge propagation over time.

physics.soc-ph↗

arXiv2015

APFEL Web: a web-based application for the graphical visualization of parton distribution functions

We present APFEL Web, a web-based application designed to provide a flexible user-friendly tool for the graphical visualization of parton distribution functions (PDFs). In this note we describe the technical design of the APFEL Web application, motivating the choices and the framework used for the development of this project. We document the basic usage of APFEL Web and show how it can be used to provide useful input for a variety of collider phenomenological studies. Finally we provide some examples showing the output generated by the application.

hep-ph↗