Dario Loi: publications

A Metacognitive Framework for Test-Time Adaptation: Taxonomy and Research Challenges

Thu, 05 Nov 2026 00:00:00 GMT

AAAI 2026 Fall Symposium on Artificial Metacognition (METACOG-26), Top 3 paper. Dario Loi, Gabriele Onorato, Antonella Coviello, Shahriar Rifat, Nathaniel D. Bastian, Francesco Restuccia.

In Test-Time Adaptation (TTA), when out-of-distribution samples interleave unpredictably with corrupted in-distribution samples, adaptation without corrective measures is prone to significant performance decrease. Existing protocols rely on unreliable factors (e.g., predictive entropy) which are unable to distinguish corrupted in-distribution data from semantically novel open-set samples. To address this issue, we formalize TTA through a two-level metacognitive framework that decouples task execution from control, facilitating a unified taxonomy that categorizes existing TTA protocols across three core axes, i.e., which knowledge is monitored, which part of the models are targeted for adaptation, and how adaptation occurs. We provide evidence that robust corrective measures in TTA require monitoring grounded in structured internal representations to reliably estimate model competence. Lastly, we frame monitoring as a resource-allocation problem, examine neuromorphic substrates for bounded self-regulation, and outline key open challenges.

Implicit Sheaf Operators for Effective and Efficient Cellular Sheaf Learning

Tue, 04 Aug 2026 00:00:00 GMT

Preprint, August 2026. Gabriele Onorato, Dario Loi, Cristian Apostol, Nathaniel D. Bastian, Francesco Restuccia.

While Sheaf Neural Networks (SNNs) enrich $n$-node $m$-edge diffusion of $f$-dimensional node features with $d$-dimensional node and edge stalks, they require computing a sparse $nd \times nd$ sheaf Laplacian, whose computation is prohibitive on edge devices. This paper addresses this issue with Implicit Sheaf Operators (ISOs), a type of sheaf topological operators that leverage the topological structure of cellular sheaves to decompose the global diffusion process into local operator-vector products. We mathematically prove that ISO eliminates the $O(md^2 \log(md^2))$ sparse-coalescing term and reduces the space complexity by $O(\min(d, f))$ with respect to traditional Neural Sheaf Diffusion (NSD). Our experiments indicate that ISO-based operators achieve latency reduction of up to 4.87× and save up to 3.92× peak GPU memory. On real graph classification data, ISO uses up to 1.24× less energy and 3.08× less on-device memory as measured on a Jetson Orin Nano. Finally, to bring ISO to practice, we introduce Sheaf Convolutional Networks (SCNs), an ISO-based SNN attaining up to 3.93% higher accuracy compared to NSD-based SNNs on node classification benchmarks.

Implicit Inversion turns CLIP into a Decoder

Thu, 23 Apr 2026 00:00:00 GMT

The Fourteenth International Conference on Learning Representations (ICLR 2026), Poster. Antonio D'Orazio, Maria Rosaria Briglia, Donato Crisostomi, Dario Loi, Emanuele Rodolà, Iacopo Masi.

CLIP is a discriminative model trained to align images and text in a shared embedding space. Due to its multimodal structure, it serves as the backbone of many generative pipelines, where a decoder is trained to map from the shared space back to images. We show that image synthesis is nevertheless possible using CLIP alone—without a pre-trained generative decoder or CLIP tuning. Our approach optimizes a frequency-aware implicit neural representation that encourages coarse-to-fine generation by stratifying frequencies across network layers. To stabilize this inverse mapping, we introduce adversarially robust initialization, a lightweight Orthogonal Procrustes projection to align local text and image embeddings, and a blending loss that anchors outputs to natural image statistics. With CLIP frozen, this framework unlocks capabilities such as text-to-image generation, style transfer, and image reconstruction. Our findings suggest that discriminative models may hold untapped generative potential, hidden in plain sight.

Topological Deep Learning Challenge 2025: Expanding the Data Landscape

Mon, 01 Dec 2025 00:00:00 GMT

1st Conference on Topology, Algebra, and Geometry in Data Science (TAG-DS 2025). Guillermo Bernárdez, Lev Telyatnikov, Mathilde Papillon, Marco Montagna, Raffael Theiler, Louisa Cornelis, Johan Mathe, Miquel Ferriol, Pavlo Vasylenko, Jan-Willem Van Looy, Lucia Testa, Bruno Neri, Donatella Genovese, Melanie Weber, Amaury Wei, Alessio Devoto, Alexander Weers, Robert Jankowski, Loris Cino, David Leko, Michael Banf, Jonas Müller, Thomas Grapentin, Taejin Paik, Abhijeet Dutta, Hugo Walter, Thomas Vaitses Fontanari, Ali Ghasemi, Dario Loi, Haitz Sáez de Ocáriz Borde, Gabriela Aguilar-Argüello, Giovanni B. da Rosa, Théo Saulus, Eric Rubiel Dolores-Cuenca, Leonardo Di Nino, Pierrick Leroy, Mario Edoardo Pandolfo, Andrea Cavallo, Yu Qin, Pavel Snopov, Amirreza Akbari, Ixchel Meza-Chávez, Louis Van Langendonck, Jared Able, Maria Yuffa Meshcheryakova, Henry Tsay, Luka Benić, Dominik Filipiak, Patrick Liu, Huidong Liang, Alexsandro Santos da Rosa Jr., Tiziana Cattai, Henrique M. Borges, Enrico Grimaldi, Manuel Lecha, Claudio Battiloro, Xuan-Chen Liu, Raj Deshpande, Graham Johnson, Igor Morgunov, Hugo Micheron, Rémi Devaux, Antoine Jardin, Tegan Emerson, Olga Fink, Nina Miolane.

This paper describes the 2025 edition of the Topological Deep Learning Challenge: Expanding the Data Landscape, hosted at the first Topology, Algebra, and Geometry in Data Science (TAG-DS) Conference. This year’s challenge aimed to address the data bottleneck in the field by systematically expanding the ecosystem of Topological Deep Learning (TDL). Powered by TopoBench, the challenge was organized into two primary missions: enriching the data landscape with diverse datasets, and advancing core data infrastructure. In particular, participants were invited to contribute to the open-source platform by implementing new dataset loaders, designing new benchmark tasks, or engineering robust, scalable data pipelines. The initiative successfully yielded 44 qualifying submissions. This paper outlines the scope of the competition and summarizes the key results and findings, highlighting the new resources now available to the TDL community.

AutoBench: Automating LLM Evaluation through Reciprocal Peer Assessment

Sun, 26 Oct 2025 00:00:00 GMT

Preprint, October 2025. Dario Loi, Elena Maria Muià, Federico Siciliano, Giovanni Trappolini, Vincenzo Crisà, Peter Kruger, Fabrizio Silvestri.

We present AutoBench, a fully automated and self-sustaining framework for evaluating Large Language Models (LLMs) through reciprocal peer assessment. This paper provides a rigorous scientific validation of the AutoBench methodology, originally developed as an open-source project by eZecute S.R.L. Unlike static benchmarks that suffer from test-set contamination and limited adaptability, AutoBench dynamically generates novel evaluation tasks while models alternately serve as question generators, contestants, and judges across diverse domains. An iterative weighting mechanism amplifies the influence of consistently reliable evaluators, aggregating peer judgments into consensus-based rankings that reflect collective model agreement. Our experiments demonstrate strong correlations with established benchmarks including MMLU-Pro and GPQA (respectively 78% and 63%), validating this peer-driven evaluation paradigm. The multi-judge design significantly outperforms single-judge baselines, confirming that distributed evaluation produces more robust and human-consistent assessments. AutoBench offers a scalable, contamination-resistant alternative to static benchmarks for the continuous evaluation of evolving language models.