Predicting and analyzing memorization within fine-tuned Large Language Models

Autor: Dentan, Jérémie, Buscaldi, Davide, Shabou, Aymen, Vanier, Sonia

Large Language Models have received significant attention due to their abilities to solve a wide range of complex tasks. However these models memorize a significant proportion of their training data, posing a serious threat when disclosed at inferenc

Externí odkaz: http://arxiv.org/abs/2409.18858

Evaluating Adversarial Robustness on Document Image Classification

Autor: Fronteau, Timothée, Paran, Arnaud, Shabou, Aymen

Adversarial attacks and defenses have gained increasing interest on computer vision systems in recent years, but as of today, most investigations are limited to images. However, many artificial intelligence models actually handle documentary data, wh

Externí odkaz: http://arxiv.org/abs/2304.12486

EfficientQA : a RoBERTa Based Phrase-Indexed Question-Answering System

Autor: Chaybouti, Sofian, Saghe, Achraf, Shabou, Aymen

State-of-the-art extractive question answering models achieve superhuman performances on the SQuAD benchmark. Yet, they are unreasonably heavy and need expensive GPU computing to answer questions in a reasonable time. Thus, they cannot be used for re

Externí odkaz: http://arxiv.org/abs/2101.02157

VisualWordGrid: Information Extraction From Scanned Documents Using A Multimodal Approach

Autor: Kerroumi, Mohamed, Sayem, Othmane, Shabou, Aymen

We introduce a novel approach for scanned document representation to perform field extraction. It allows the simultaneous encoding of the textual, visual and layout information in a 3-axis tensor used as an input to a segmentation model. We improve t

Externí odkaz: http://arxiv.org/abs/2010.02358

Coding of tag models

Autor: Znaidia, Amel, Shabou, Aymen, Le Borgne, Hervé, Hudelot, Céline

Publikováno v: Revue des Sciences et Technologies de l'Information-Série RIA : Revue d'Intelligence Artificielle
Revue des Sciences et Technologies de l'Information-Série RIA : Revue d'Intelligence Artificielle, 2013, 27 (1), pp.39-63. ⟨10.3166/ria.27.39-63⟩
Revue des Sciences et Technologies de l'Information-Série RIA : Revue d'Intelligence Artificielle, Lavoisier, 2013, 27 (1), pp.39-63. ⟨10.3166/ria.27.39-63⟩

National audience; Annotating images using a fixed number of concepts is a fundamental task for content based image retrieval and classification. In practice, several modalities (visual, text...) provide information about the content of images. We ar

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=dedup_wf_001::08e99763c66bd0957b8a1569b7ea4ba6
https://hal.science/hal-00824580