arXiv subjects

Andrea Favalli

Publications and source records attributed to Andrea Favalli.

9 recordsLinked to original sources

arXiv2025

A Fast, Accurate and Oscillation-free Spectral Collocation Solver for High-dimensional Transport Problems

Transport phenomena-describing the movement of particles, energy, or other physical quantities-are fundamental in various scientific disciplines, including nuclear physics, plasma physics, astrophysics, engineering, and the natural sciences. However, solving the associated seven-dimensional transport equations poses a significant computational challenge due to the curse of dimensionality. We introduce the Tensor Train Superconsistent Spectral (T${^2}$S${^2}$) solver to address this challenge, integrating Spectral Collocation for exponential convergence, Superconsistency for stabilization in transport-dominated regimes, and Tensor Train format for substantial data compression. T${^2}$S${^2}$ enforces a dimension-wise superconsistent condition compatible with tensor structures, achieving extremely low compression ratios, in the order of $(10^{-12})$, while preserving spectral accuracy. Numerical experiments on linear problems demonstrate that T${^2}$S${^2}$ can solve high-dimensional transport problems in minutes on standard hardware, making previously intractable problems computationally feasible. This advancement opens new avenues for efficiently and accurately modeling complex transport phenomena.

math.NA↗

arXiv2025

MeMo: Towards Language Models with Associative Memory Mechanisms

Memorization is a fundamental ability of Transformer-based Large Language Models, achieved through learning. In this paper, we propose a paradigm shift by designing an architecture to memorize text directly, bearing in mind the principle that memorization precedes learning. We introduce MeMo, a novel architecture for language modeling that explicitly memorizes sequences of tokens in layered associative memories. By design, MeMo offers transparency and the possibility of model editing, including forgetting texts. We experimented with the MeMo architecture, showing the memorization power of the one-layer and the multi-layer configurations.

cs.CL↗

arXiv2024

Enhancing Data Privacy in Large Language Models through Private Association Editing

Large language models (LLMs) require a significant redesign in solutions to preserve privacy in data-intensive applications due to their text-generation capabilities. Indeed, LLMs tend to memorize and emit private information when maliciously prompted. In this paper, we introduce Private Association Editing (PAE) as a novel defense approach for private data leakage. PAE is designed to effectively remove Personally Identifiable Information (PII) without retraining the model. Experimental results demonstrate the effectiveness of PAE with respect to alternative baseline methods. We believe PAE will serve as a critical tool in the ongoing effort to protect data privacy in LLMs, encouraging the development of safer models for real-world applications.

cs.CL↗

arXiv2024

Investigating the Impact of Data Contamination of Large Language Models in Text-to-SQL Translation

Understanding textual description to generate code seems to be an achieved capability of instruction-following Large Language Models (LLMs) in zero-shot scenario. However, there is a severe possibility that this translation ability may be influenced by having seen target textual descriptions and the related code. This effect is known as Data Contamination. In this study, we investigate the impact of Data Contamination on the performance of GPT-3.5 in the Text-to-SQL code-generating tasks. Hence, we introduce a novel method to detect Data Contamination in GPTs and examine GPT-3.5's Text-to-SQL performances using the known Spider Dataset and our new unfamiliar dataset Termite. Furthermore, we analyze GPT-3.5's efficacy on databases with modified information via an adversarial table disconnection (ATD) approach, complicating Text-to-SQL tasks by removing structural pieces of information from the database. Our results indicate a significant performance drop in GPT-3.5 on the unfamiliar Termite dataset, even with ATD modifications, highlighting the effect of Data Contamination on LLMs in Text-to-SQL translation tasks.

cs.CL↗

arXiv2023

Vanquishing the computational cost of passive gamma emission tomography simulations: a physics-aware reduced order modeling approach

Passive Gamma Emission Tomography (PGET) has been developed by the International Atomic Energy Agency as a way to directly image the spatial distribution of individual fuel pins in a spent nuclear fuel assembly and so determine potential diversion. Constructing the analysis and interpretation of PGET measurements rely on the availability of comprehensive datasets. Experimental data are expensive, limited, and so are augmented by Monte Carlo simulations. The main issue concerning Monte Carlo simulations is the high computational cost to simulate the 360 angular views of the tomography. Similar challenges pervade numerical science. To address this challenge, we have developed a physics-aware reduced order modeling approach. It provides a framework to combine a small subset of the 360 angular views with a computationally inexpensive proxy solution, that brings the essence of the physics, to obtain a real-time high-fidelity solution at all angular views, but at a fraction of the computational cost.

math.NA↗

arXiv2021

Every time I fire a conversational designer, the performance of the dialog system goes down

Incorporating explicit domain knowledge into neural-based task-oriented dialogue systems is an effective way to reduce the need of large sets of annotated dialogues. In this paper, we investigate how the use of explicit domain knowledge of conversational designers affects the performance of neural-based dialogue systems. To support this investigation, we propose the Conversational-Logic-Injection-in-Neural-Network system (CLINN) where explicit knowledge is coded in semi-logical rules. By using CLINN, we evaluated semi-logical rules produced by a team of differently skilled conversational designers. We experimented with the Restaurant topic of the MultiWOZ dataset. Results show that external knowledge is extremely important for reducing the need of annotated examples for conversational systems. In fact, rules from conversational designers used in CLINN significantly outperform a state-of-the-art neural-based dialogue system.

cs.CL↗

arXiv2019

Almawave-SLU: A new dataset for SLU in Italian

The widespread use of conversational and question answering systems made it necessary to improve the performances of speaker intent detection and understanding of related semantic slots, i.e., Spoken Language Understanding (SLU). Often, these tasks are approached with supervised learning methods, which needs considerable labeled datasets. This paper presents the first Italian dataset for SLU. It is derived through a semi-automatic procedure and is used as a benchmark of various open source and commercial systems.

cs.CL↗

arXiv2019

Multi-lingual Intent Detection and Slot Filling in a Joint BERT-based Model

Intent Detection and Slot Filling are two pillar tasks in Spoken Natural Language Understanding. Common approaches adopt joint Deep Learning architectures in attention-based recurrent frameworks. In this work, we aim at exploiting the success of "recurrence-less" models for these tasks. We introduce Bert-Joint, i.e., a multi-lingual joint text classification and sequence labeling framework. The experimental evaluation over two well-known English benchmarks demonstrates the strong performances that can be obtained with this model, even when few annotated data is available. Moreover, we annotated a new dataset for the Italian language, and we observed similar performances without the need for changing the model.

cs.CL↗

arXiv2016

Experimental observation of $β$-delayed neutrons from $^{9}$Li as a way to study short-pulse laser-driven deuteron production

A short-pulse laser-driven deuteron beam is generated in the relativistic transparency regime and aimed at a beryllium converter to generate neutrons at the TRIDENT laser facility. These prompt neutrons have been used for active interrogation to detect nuclear materials, the first such demonstration of a laser-driven neutron source. During the experiments, delayed neutrons from $^9$Li decay was observed. It was identified by its characteristic half-life of 178.3 ms. Production is attributed to the nuclear reactions $^9$Be(d,2p)$^9$Li and $^9$Be(n,p)$^9$Li inside the beryllium converter itself. These reactions have energy thresholds of 18.42 and 14.26 MeV respectively, and we estimate the (d,2p) reaction to be the dominant source of $^9$Li production. Therefore, only the higher-energy portion of the deuteron spectrum contributes to the production of the delayed neutrons. It was observed that the delayed-neutron yield decreases with increasing distance between the converter and the deuteron source. This behavior is consistent with deuteron production with energy greater than $\sim$20 MeV within a cone with a half-angle greater than 40$^{\circ}$. Prompt-neutron time-of-flight measurements at varying separation between the converter and the laser target indicate that the fast deuteron population above threshold is severely depleted on axis out to $\sim$20$^{\circ}$. These measurements are consistent with emission of the fast deuterons (i.e., above 10 MeV/nucleon) in a ring-like fashion around the central axis. Such an inferred ring-like structure is qualitatively consistent with a documented signature of the breakout afterburner (BOA) laser-plasma ion acceleration mechanism. The measurement of $β$-delayed neutrons from $^9$Li decay could provide an important new diagnostic tool for the study of the features of the deuteron production mechanism in a non-intrusive way.

physics.plasm-ph↗