Gabriele Sarti

Profile

I am a PhD student at the Computational Linguistics Group of the University of Groningen and member of the InDeep consortium, currently working on user-centric interpretability of large language models. My supervisors are Arianna Bisazza, Malvina Nissim and Grzegorz Chrupała.

Previously, I was a research intern at Amazon Translate NYC, a research scientist at Aindo, a Data Science MSc student at the University of Trieste and a co-founder of the AI Student Society.

My research focuses on bridging the gap between advances in the field of interpretability for generative language models and the downstream benefits to model users, with a particular emphasis on understanding how contextual information is integrated into predictions to improve model trustworthiness. I am also very interested in parallels between human and artificial learning and reasoning, with a particular taste for working with human behavioral signals.

I am the main developer of the Inseq library for LLM interpretability, and I am generally very excited about open-source projects making interpretability tools and techniques more accessible to the broader AI community.

Publications

Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses

Non Verbis, Sed Rebus: Large Language Models are Weak Solvers of Italian Rebuses

Gabriele Sarti, Tommaso Caselli, Malvina Nissim, Arianna Bisazza

10th Italian Conference on Computational Linguistics (CLiC-it 2024)

Multi-property Steering of Large Language Models with Dynamic Activation Composition

Multi-property Steering of Large Language Models with Dynamic Activation Composition

Daniel Scalena, Gabriele Sarti, Malvina Nissim

7th Workshop on Analyzing and Interpreting Neural Networks for NLP (BlackboxNLP 2024)

Model Internals-based Answer Attribution for Trustworthy Retrieval-Augmented Generation

Model Internals-based Answer Attribution for Trustworthy Retrieval-Augmented Generation

Jirui Qi, Gabriele Sarti, R. Fernández, Arianna Bisazza

Conference on Empirical Methods in Natural Language Processing (EMNLP 2024)

A Primer on the Inner Workings of Transformer-based Language Models

A Primer on the Inner Workings of Transformer-based Language Models

Javier Ferrando, Gabriele Sarti, Arianna Bisazza, Marta Costa-jussà

Arxiv Preprint

DecoderLens: Layerwise Interpretation of Encoder-Decoder Transformers

DecoderLens: Layerwise Interpretation of Encoder-Decoder Transformers

Anna Langedijk, Hosein Mohebbi, Gabriele Sarti, Willem Zuidema, Jaap Jumelet

Findings of the Association for Computational Linguistics: NAACL 2024

Quantifying the Plausibility of Context Reliance in Neural Machine Translation

Quantifying the Plausibility of Context Reliance in Neural Machine Translation

Gabriele Sarti, Grzegorz Chrupala, Malvina Nissim, Arianna Bisazza

International Conference on Learning Representations (ICLR 2024)

RAMP: Retrieval and Attribute-Marking Enhanced Prompting for Attribute-Controlled Translation

RAMP: Retrieval and Attribute-Marking Enhanced Prompting for Attribute-Controlled Translation

Gabriele Sarti, Phu Mon Htut, Xing Niu, Benjamin Hsu, Anna Currey, Georgiana Dinu, Maria Nadejde

Annual Meeting of the Association for Computational Linguistics (ACL 2023)

Are Character-level Translations Worth the Wait? Comparing ByT5 and mT5 for Machine Translation

Are Character-level Translations Worth the Wait? Comparing ByT5 and mT5 for Machine Translation

Lukas Edman, Gabriele Sarti, Antonio Toral, Gertjan van Noord, Arianna Bisazza

Transactions of the Association for Computational Linguistics (2024) 12: 392–410

Inseq: An Interpretability Toolkit for Sequence Generation Models

Inseq: An Interpretability Toolkit for Sequence Generation Models

Gabriele Sarti, Nils Feldhus, Ludwig Sickert, Oskar van der Wal, Malvina Nissim, Arianna Bisazza

Annual Meeting of the Association for Computational Linguistics (ACL 2023): Demo Track

Probing Linguistic Knowledge in Italian Neural Language Models across Language Varieties

Probing Linguistic Knowledge in Italian Neural Language Models across Language Varieties

Alessio Miaschi, Gabriele Sarti, Dominique Brunato, Felice Dell’Orletta, Giulia Venturi

Italian Journal of Computational Linguistics (IJCoL)

DivEMT: Neural Machine Translation Post-Editing Effort Across Typologically Diverse Languages

DivEMT: Neural Machine Translation Post-Editing Effort Across Typologically Diverse Languages

Gabriele Sarti, Arianna Bisazza, Ana Guerberof Arenas, Antonio Toral

Conference on Empirical Methods in Natural Language Processing (EMNLP 2022)

IT5: Text-to-text Pretraining for Italian Language Understanding and Generation

IT5: Text-to-text Pretraining for Italian Language Understanding and Generation

Gabriele Sarti, Malvina Nissim

International Conference on Language Resources and Evaluation, International Conference on Computational Linguistics (LREC-COLING 2024)

Contrastive Language-Image Pre-training for the Italian Language

Contrastive Language-Image Pre-training for the Italian Language

Federico Bianchi, Giuseppe Attanasio, Raphael Pisoni, Silvia Terragni, Gabriele Sarti

9th Italian Conference on Computational Linguistics (CLiC-it 2023)

That Looks Hard: Characterizing Linguistic Complexity in Humans and Language Models

That Looks Hard: Characterizing Linguistic Complexity in Humans and Language Models

Gabriele Sarti, Dominique Brunato, Felice Dell’Orletta

Workshop on Cognitive Modeling and Computational Linguistics (CMCL 2021)

Teaching NLP with Bracelets and Restaurant Menus: An Interactive Workshop for Italian Students

Teaching NLP with Bracelets and Restaurant Menus: An Interactive Workshop for Italian Students

Ludovica Pannitto, Lucia Busso, Claudia Roberta Combei, Lucio Messina, Alessio Miaschi, Gabriele Sarti, Malvina Nissim

Workshop on Teaching NLP 2021

UmBERTo-MTSA @ AcCompl-It: Improving Complexity and Acceptability Prediction with Multi-task Learning on Self-Supervised Annotations (short paper)

UmBERTo-MTSA @ AcCompl-It: Improving Complexity and Acceptability Prediction with Multi-task Learning on Self-Supervised Annotations (short paper)

Gabriele Sarti

International Workshop on Evaluation of Natural Language and Speech Tools for Italian 2020

ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation

ETC-NLG: End-to-end Topic-Conditioned Natural Language Generation

Ginevra Carbone, Gabriele Sarti

Italian Journal of Computational Linguistics (IJCoL)

Democratizing Advanced Attribution Analyses of Generative Language Models with the Inseq Toolkit

Democratizing Advanced Attribution Analyses of Generative Language Models with the Inseq Toolkit

Gabriele Sarti, Nils Feldhus, Jirui Qi, Malvina Nissim, Arianna Bisazza

xAI 2024

Italian Transformers Under the Linguistic Lens

Italian Transformers Under the Linguistic Lens

Alessio Miaschi, Gabriele Sarti, D. Brunato, F. Dell’Orletta, Giulia Venturi

7th Italian Conference on Computational Linguistics (CLiC-it 2020)

ArchiMeDe @ DANKMEMES: A New Model Architecture for Meme Detection

ArchiMeDe @ DANKMEMES: A New Model Architecture for Meme Detection

Jinen Setpal, Gabriele Sarti

Workshop on Evaluation of Natural Language and Speech Tools for Italian (EVALITA 2020)