Autonomous Linguistic Emergence in neural Networks

Informazioni relative al progetto

ALiEN

ID dell’accordo di sovvenzione: 101019291

DOI

10.3030/101019291

Data della firma CE 26 Maggio 2021

Data di avvio 1 Gennaio 2022

Data di completamento 31 Dicembre 2026

Finanziato da

• EXCELLENT SCIENCE - European Research Council (ERC)

Costo totale

€ 2 489 541,00

Contributo UE

€ 2 489 541,00

Coordinato da

UNIVERSIDAD POMPEU FABRA
Spain

CORDIS fornisce collegamenti ai risultati finali pubblici e alle pubblicazioni dei progetti ORIZZONTE.

I link ai risultati e alle pubblicazioni dei progetti del 7° PQ, così come i link ad alcuni tipi di risultati specifici come dataset e software, sono recuperati dinamicamente da .OpenAIRE .

Pubblicazioni

Conference proceedings (19)

The emergence of numerical representations in communicating artificial agents

Autori: Daniela Mihai, Lucas Weber, Francesca Franzon
Pubblicato in: Proceedings of Evolang XVI, 2026, Pagina/e 308-315
Editore: The Evolution of Language Conferences

Cross-domain image captioning with discriminative finetuning.

Autori: R. Dessì, M. Bevilacqua, E. Gualdoni, N. Rakotonirina, F. Franzon, M. Baroni
Pubblicato in: Proceedings of CVPR 2023 (IEEE/CVF Conference on Computer Vision and Pattern Recognition), 2023, Pagina/e 6935-6944
Editore: Computer Vision Foundation

Towards hate speech detection in low-resource languages: Comparing ASR to acoustic word embeddings on Wolof and Swahili

Autori: Jacobs, C., Rakotonirina, N.C., Chimoto, E.A., Bassett, B.A., Kamper, H.
Pubblicato in: Proc. INTERSPEECH 2023, 2023
Editore: ISCA
DOI: 10.21437/interspeech.2023-421

On the Correspondence between Compositionality and Imitation in Emergent Neural Communication

Autori: E. Cheng, M. Rita, T. Poibeau
Pubblicato in: Findings of the Association for Computational Linguistics: ACL 2023, 2023, Pagina/e 12432–12447
Editore: ACL

Referential communication in heterogeneous communities of pre-trained visual deep networks.

Autori: M. Mahaut, F. Franzon, R. Dessì, M. Baroni
Pubblicato in: Proc. of the 22nd International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2023), 2023, Pagina/e 2619-2621
Editore: International Foundation for Autonomous Agents and Multiagent Systems

Not a nuisance but a useful heuristic: Outlier dimensions favor frequent tokens in language models

Autori: Iuri Macocco, Nora Graichen, Gemma Boleda, Marco Baroni
Pubblicato in: Proceedings of the 8th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP, 2025, Pagina/e 109-136
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2025.blackboxnlp-1.6

Geometric Signatures of Compositionality Across a Language Model’s Lifetime

Autori: Jin Hwa Lee, Thomas Jiralerspong, Lei Yu, Yoshua Bengio, Emily Cheng
Pubblicato in: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025, Pagina/e 5292-5320
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2025.acl-long.265

Communication breakdown: On the low mutual intelligibility between human and neural captioning

Autori: R. Dessì, E. Gualdoni, F. Franzon, G. Boleda, M. Baroni
Pubblicato in: Proceedings of EMNLP 2022 (Conference on Empirical Methods in Natural Language Processing), 2022, Pagina/e 7998-8007
Editore: ACL

Factual Confidence of LLMs: on Reliability and Robustness of Current Estimators

Autori: Matéo Mahaut, Laura Aina, Paula Czarnowska, Momchil Hardalov, Thomas Müller, Lluis Marquez
Pubblicato in: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2024, Pagina/e 4554-4570
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2024.acl-long.250

Evil twins are not that evil: Qualitative insights into machine-generated prompts

Autori: Nathanaël Carraz Rakotonirina, Corentin Kervadec, Francesca Franzon, Marco Baroni
Pubblicato in: Proceedings of the 8th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP, 2025, Pagina/e 48-68
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2025.blackboxnlp-1.3

Bridging information-theoretic and geometric compression in language models

Autori: Cheng, Emily; Kervadec, Corentin; Baroni, Marco
Pubblicato in: Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023
Editore: Association for Computational Linguistics

MemoryPrompt: A light wrapper to improve context tracking in pre-trained language models

Autori: Rakotonirina, Nathanael; Baroni, Marco
Pubblicato in: Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), 2024
Editore: Association for Computational Linguistics

Evidence from fMRI Supports a Two-Phase Abstraction Process in Language Models

Autori: Richard Antonello, Emily Cheng
Pubblicato in: Proceedings of UniReps@NeurIPS 2024, 2024
Editore: OpenReview/Online

Prediction Hubs are Context-Informed Frequent Tokens in LLMs

Autori: Beatrix Miranda Ginn Nielsen, Iuri Macocco, Marco Baroni
Pubblicato in: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025, Pagina/e 23715-23745
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2025.acl-long.1156

Unnatural language processing: How do language models handle machine-generated prompts?

Autori: Kervadec, Corentin; Franzon, Francesca; Baroni, Marco
Pubblicato in: Findings of EMNLP, 2023
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2023.findings-emnlp.959

From Tools to Teammates: Evaluating LLMs in Multi-Session Coding Interactions

Autori: Nathanaël Carraz Rakotonirina, Mohammed Hamdy, Jon Ander Campos, Lucas Weber, Alberto Testoni, Marzieh Fadaee, Sandro Pezzelle, Marco Del Tredici
Pubblicato in: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025, Pagina/e 19609-19642
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2025.acl-long.964

Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning

Autori: Nathanaël Carraz Rakotonirina, Ren Pang, Neha Anna John, Michael Bohlke-Schneider, Momchil Hardalov
Pubblicato in: Findings of the Association for Computational Linguistics: ACL 2026, 2026, Pagina/e 12798-12810
Editore: Association for Computational Linguistics
DOI: 10.18653/v1/2026.findings-acl.622

Can discrete information extraction prompts generalize across language models?

Autori: N. Rakotonirina, R. Dessì, F. Petroni, S. Riedel, M. Baroni
Pubblicato in: Proceedings of ICLR 2023 (International Conference on Learning Representations), 2023
Editore: ICLR

Emergence of a high-dimensional abstraction phase in language transformers

Autori: Emily Cheng, Diego Doimo, Corentin Kervadec, Iuri Macocco, Jade Yu, Alessandro Laio and Marco Baroni
Pubblicato in: Proceedings of ICLR 2025, 2025
Editore: ICLR

Peer reviewed articles (2)

LiSeCo: Linear Semantic Control for Language Generation

Autori: Emily Cheng, Carmen Amo Alonso
Pubblicato in: Transactions of Machine Learning Research, 2026, ISSN 2835-8856
Editore: Journal of Machine Learning Research Inc

Referential communication in heterogeneous communities of pre-trained visual deep networks

Autori: Mateo Mahaut, Francesca Franzon, Roberto Dessi, Marco Baroni
Pubblicato in: Transactions on Machine Learning Research, 2025, ISSN 2835-8856
Editore: Journal of Machine Learning Research Inc