Description
TALEVA is a joint project between TALIA Srl and CNR-ILC, carried out as industrial research and experimental development, whose goal is an advanced platform for evaluating the performance of Large Language Models. The project responds to a growing need for reliable, scientifically grounded and reproducible instruments for comparing and validating LLMs, whose adoption has accelerated with the spread of open-source models that can be run on widely available hardware. TALEVA addresses the structural lack of standardised evaluation methodologies that would allow companies, institutions and research centres to select, customise and adopt these models in an informed way.
The platform integrates heterogeneous evaluation approaches, combining lexical and semantic metrics, model-based judging, and controlled human annotation, so that results are transparent, comparable and replicable. Particular attention is devoted to Italian, where the project aims to narrow the gap with respect to international benchmarks. The outcome will be a fully functioning prototype, tested in realistic use contexts, designed to demonstrate technical reliability, operational effectiveness and potential for transfer to the market.
Duration: October 2026 — January 2028 · Type: Regional project (Regione Toscana, PR FESR 2021/2027) · Coordinator: TALIA Srl