arXiv subjects

Davide Cifarelli

Publications and source records attributed to Davide Cifarelli.

3 recordsLinked to original sources

arXiv2025

A new method for structural diagnostics with muon tomography and deep learning

This work investigates the production of high-resolution images of typical support elements in concrete structures by means of muon tomography (muography). By exploiting detailed Monte Carlo radiation-matter simulations, we demonstrate the feasibility of reconstructing 1 cm-thick iron bars inside 30 cm-deep concrete blocks, regarded as an important testbed within the structural diagnostics community. In addition, we present a new method for integrating simulated data with advanced deep learning techniques in order to improve the muon imaging of concrete structures. Through deep learning enhancement techniques, this results in a dramatic improvement in image quality and a significant reduction in data acquisition time, which are two critical limitations within the usual practice of muography for civil engineering diagnostics.

hep-ex↗

arXiv2023

Safurai-Csharp: Harnessing Synthetic Data to improve language-specific Code LLM

This paper introduces Safurai-Csharp, an open-source model designed to specialize in the generation, completion, and debugging of C# code. Safurai-Csharp is built upon the novel CodeLlama 34B model and leverages the EvolInstruct technique, creating a refined and expanded dataset for its fine-tuning process. The results of its performance, a notable score of 56.33% on the Manual MultiPL-E benchmark (Zero-Shot, Pass@1), signal its high capacity to streamline developers' workflows and aid code learning. It shows promise in setting new stakes in the landscape of open-source C# LLMs and hopes to inspire more inclusive and wide-ranging development in the field of language-specific LLMs.

cs.CL↗

arXiv2023

Safurai 001: New Qualitative Approach for Code LLM Evaluation

This paper presents Safurai-001, a new Large Language Model (LLM) with significant potential in the domain of coding assistance. Driven by recent advancements in coding LLMs, Safurai-001 competes in performance with the latest models like WizardCoder [Xu et al., 2023], PanguCoder [Shen et al., 2023] and Phi-1 [Gunasekar et al., 2023] but aims to deliver a more conversational interaction. By capitalizing on the progress in data engineering (including latest techniques of data transformation and prompt engineering) and instruction tuning, this new model promises to stand toe-to-toe with recent closed and open source developments. Recognizing the need for an efficacious evaluation metric for coding LLMs, this paper also introduces GPT4-based MultiParameters, an evaluation benchmark that harnesses varied parameters to present a comprehensive insight into the models functioning and performance. Our assessment shows that Safurai-001 can outperform GPT-3.5 by 1.58% and WizardCoder by 18.78% in the Code Readability parameter and more.

cs.CL↗