Источник
ACL / GenBench
Дата публикации
16.11.2024
Авторы
Александр Панченко Антон Разжигаев Андрей Кузнецов Денис Димитров Елизавета Гончарова Максим Куркин Ирина Абдуллаева Анастасия Лысенко
Поделиться

OmniGen: A Multimodal Benchmark for Generalization Across Text, Visual, and Audio Modalities

Аннотация

We introduce OmniDialog — the first trimodal comprehensive benchmark grounded in a knowledge graph (Wikidata) to evaluate the generalization of Large Multimodal Models (LMMs) across three modalities. Our benchmark consists of more than 4,000 dialogues, each averaging 10 turns, all annotated and cross-validated by human experts. The dialogues in our dataset are designed to prevent shortcut learning by incorporating various formats and misleading or irrelevant multimodal cues. We also evaluate both multimodal and unimodal models to gain insights into how they process modality inputs introduced in the conversation.

Присоединяйтесь к AIRI в соцсетях