CIIT Lab UNISA

Today the CIIT Lab UNISA participated to the first edition of the showcasing some of the lab projects and activities in the area of , and

Thanks to all the companies and visitors that came to visit us!



Università degli Studi di Salerno Comunicazione Unisa

Yesterday the paper "MIRAGE: A Multi-Agent, Multi-RAG, Architecture for Metaphor-Driven Advertising Narratives and Commercial Video Generation" by Antonio Lieto and Anjali Sneha Kujur was presented at the WOA 2026, the 27th Workshop From Objects to Agents.

The paper presents a multi-agent and multi-rag pipeline able to generate, starting from a simple list of product features sent via a Telegram bot, an advertising video leveraging a metaphors-driven storytelling strategy.

The core component of the pipeline is represented by the Metaphor-RAG that employs the output of the MetCL system presented last year at IJCAI https://lnkd.in/dkMBbxv

MIRAGE (Metaphor-Informed Retrieval-Augmented Generative Engine) was developed during the internship of Anjali Sneha Kujur at the CIIT Lab, Cognition Interaction and Intelligent Technologies Laboratory.

Abstract:
Early-stage product design often begins with short, informal, descriptions that are rich in intent but poor in structure. Converting such descriptions into persuasive communication assets typically requires multiple human roles, including product strategists, copywriters, and creative directors. This paper presents a multi-agent, multi-RAG architecture - called MIRAGE (Metaphor-Informed Retrieval-Augmented Generative Engine) - that automates this transformation from a Telegram brief to a metaphor-grounded marketing story and finally to a self-explanatory commercial video. The central design choice of the system is to use metaphor as an explicit intermediate element used to generate a product-oriented story script upon which the video is generated. The proposed architecture offers a modular approach to controllable creative generation in product design settings. We report also the result of a preliminary user study evaluating the generated outcome and indicate a number of possibile research direction.

Preprint:https://www.ciitlab.org/mirage-woa2026.pdf
YouTube playlist with the generated videos (avatar-talking scenario and commercials): https://www.youtube.com/playlist?list=PL3QJL5trfgE9Wlcvj0jECQ5MNNeN7uuW-

Today at the CIIT Lab UNISA (https://www.ciitlab.org) we have hosted the seminar of Dr. Antonio Schiano, Camilla Giacometti and Angela Caiazza of NTT DATA entitled "What is Agentic AI, use cases and risks".

Lot of questions and interests from students and a great collaboration ahead with shared interns and projects!



Università degli Studi di Salerno CIIT Lab UNISA Comunicazione Unisa

On May 27, as part of the CIIT Lab UNISA activities, a seminar entitled "What Is Agentic AI, Use Cases and Risks" will be held in Room 021 of Building B2 from 10:30 a.m. to 12:30 p.m.
https://www.ciitlab.org/invited-seminar-by-dr-antonio-schiano-ntt-data/

The seminar—open to all and held as part of the integrative activities of the Policom doctorate—will be led by Dr. Antonio Schiano, Senior Manager of NTT DATA, Inc. (a leading global software integration and AI company).

The seminar will provide an introduction to the topic of so-called "Agentic AI": multi-agent artificial intelligence that is finding application in various fields.

The Director of the CIIT Lab UNISA, prof. Antonio Lieto, will deliver tomorrow an invited talk at the Doctoral School of the Aalto University in Design, Art and Media in Helsinki on the thene: "Cognitive Design for AI systems with human-like reasoning"

Abstract: The current success of artificial intelligence (AI) systems has fueled the hypothesis that their ability to achieve human-level (or superhuman) performances in several significant tasks suggests that these systems have acquired the skills we typically attribute to humans when they are capable of
such behaviors. This type of skill attribution, however, is methodologically problematic since current mainstream AI systems are not based on a cognitive design perspective. In this talk I will introduce the framework of cognitive design for AI and will present some concrete applications in the field of commonsense reasoning.