Almawave will participate in the 63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

From July 27 to August 1, Vienna will host ACL 2025, the 63rd Annual Meeting of the Association for Computational Linguistics, one of the most important international events in the field of Natural Language Processing (NLP) and Computational Linguistics.

Almawave will be among the key participants at the event with Giancarlo Xompero, Machine Learning Engineer, and Fabio Massimo Zanzotto, Senior Research Scientist at Almawave and Associate Professor at the University of Tor Vergata, who will present two scientific papers:

• Towards Language Models with Associative Memory Mechanisms

• Private Memorization Editing: Turning Memorization into a Defense to Strengthen Data Privacy in Large Language Models

The first paper introduces MeMo, a new architecture for language modeling based on layered associative memories, enabling models to explicitly store sequences of tokens. The goal is to establish a more transparent and controllable paradigm, where memorization precedes learning and can be observed, modified, or erased.

The second paper presents Private Memorization Editing (PME), an innovative method to address the risks of unintended memorization of sensitive data in large language models (LLMs). PME proposes an approach that turns memorization into a privacy defense mechanism, allowing the model to “forget” confidential information without compromising its language performance.

Almawave reaffirms its commitment to cutting-edge research in Artificial Intelligence, contributing to the development of more ethical, transparent, and secure language models.