Paper-Conference

WEALY matches songs by their lyrics straight from the audio, using Whisper speech embeddings, and offers a reproducible baseline across three public datasets.

Twenty-four language models score Italian news summaries against an editorial rubric, and most of them rate the summaries higher than a human expert does.

A community roadmap for evaluating Italian LLMs: what to test, how to test it, where to find data, and how to keep benchmarking sustainable.

RECV, a benchmark of real-world claims labeled by reasoning type, shows that LLMs verify claims by deduction well but fail when abduction is needed.

GenSPP trains the generator and predictor of a self-explaining text classifier separately with genetic search, removing interlocking and producing more accurate highlights.

MM-ArgFallacy2025 benchmarks fallacy detection and classification in U.S. presidential debates from text, audio, or both, and finds that text-based systems still score highest.

HateDefCon breaks 493 hate speech definitions from five domains into semantic components and shows LLM hate speech detection changes with the definition in the prompt.

A pipeline trains small classifiers on LLM-labelled EU tax rulings to find the Court’s interpretative paragraphs, matching the LLM on expert-labelled test data.

A new corpus of 30 Italian privacy policies shows that BERT models beat LLMs at flagging vague data clauses and that resources transfer across languages.

Choosing few-shot examples per query by similarity lets LLMs mine legal arguments in EU court rulings as well as or better than expert-chosen examples.