arXiv subjects

Paulo Henrique Eleuterio Falsetti

Publications and source records attributed to Paulo Henrique Eleuterio Falsetti.

3 recordsLinked to original sources

arXiv2026

A Reproducible, License-Aware Distillation Recipe for CPUDeployable Safety Classification

Deploying a safety layer for large language models on commodity hardware is constrained by the guards available to do it: current open guard models hold between 1 and 9 billion parameters, are oriented toward the graphics processing unit, and answer in seconds per request on a central processing unit. This paper presents a reproducible, license-aware knowledge-distillation recipe addressing that constraint. A strong open guard labels a corpus of roughly 97,000 prompts, drawn from 24 public datasets, into seven safety categories aligned to a public hazard taxonomy, and a fleet of small students spanning lexical, shallow, encoder and generative architectures is trained to reproduce that signal. The corpus is partitioned at the license boundary, so that a deployable and a research model differ only in their training data and the cost of that restriction becomes measurable. Every model is scored against an independent gold benchmark of 6,361 rows over four slices, labeled apart from the teacher and including a slice of harmless prompts that makes over-defense measurable. The distilled students match the teachers on adversarial text within overlapping confidence intervals and reduce false alarms on harmless prompts, the smallest generative student reaching 3.8% against 4.8% for the 8-billion-parameter teacher, while the encoder classifies in roughly 24 ms per request on CPU. Per-class rebalancing is the only decisive ingredient of the recipe. No superiority over the distilled guards is claimed; on the clean reference slice they remain ahead.

cs.AI↗

arXiv2026

GuardNet: Ensemble Strategies of Shallow Neural Networks for Robust Prompt Injection and Jailbreak Detection

Large Language Models (LLMs) have transformed natural language processing, but they remain vulnerable to Prompt Injection (PI) and Jailbreak (JB) attacks. In addition, benchmark evaluations may be affected by contamination and partial information leakage, compromising performance estimates. This work presents GuardNet, a guardrail system based on an ensemble of shallow neural networks (BiLSTMs) with approximately 47 million parameters. We investigate the hypothesis that robustness in adversarial scenarios depends more on the diversity of example coverage and threshold calibration than on model scale. The results indicate that GuardNet achieves competitive performance compared with lightweight detectors and high efficiency at low latency, although larger LLMs such as Mistral-7B and Llama-3.1-8B still achieve superior performance in terms of F1 score and AUROC on the blind JBB-Behaviors benchmark. Nevertheless, GuardNet achieves an AUROC of 0.747 on the blind dataset (n = 200) and an F1 score of 0.92 on a proprietary benchmark (n = 50), under threshold calibration and evaluation with declared partial information leakage. The system operates with an average latency of approximately 50 ms on CPU, making it suitable for deployment in production environments with cost and infrastructure constraints.

cs.AI↗

arXiv2025

Estimation of nanometer-thickness layer of 6B graphite: an experimental activity to study Ohm's law in higher education

Ohm's law is crucial for understanding electrical circuits and conductive materials. Formulated by Georg Simon Ohm in the 19th century, it describes the direct proportional relationship between electric voltage and electric current in ohmic conductors. Despite its apparent simplicity, the literature has pointed out that students at different educational levels have difficulty understanding it, especially due to the abstraction associated with the concepts of electric voltage, electric current, and electrical resistance. In order to provide possibilities to overcome this abstraction and the associated learning difficulties, the present work proposes an experimental activity that applies Ohm's law to determine the nanometer-scale thickness of 6B graphite traces deposited on tracing paper. The employed methodology is based on measuring physical quantities such as electric voltage, electric current, and length, and on analyzing the collected data. The thickness of the graphite traces determined by this method was also compared with values obtained using a scanning probe microscope (SPM). Based on the proposed methodology, the thickness of the traces was determined to be 456.5(34)~nm, while SPM measurements yielded an average thickness of 440(50)~nm. Thus, the results show good agreement between the measurements within experimental uncertainties, validating the effectiveness of the suggested methodology and indicating that it is a viable proposal for electricity courses in higher education and, with adaptations, even for high school.

physics.ed-ph↗