We introduce Vibe-Eval: a new open benchmark and framework for evaluating multimodal chat models. Vibe-Eval consists of 269 visual understanding prompts, including 100 of hard difficulty, complete with gold-standard responses authored by experts. Vibe-Eval is open-ended and challenging with dual objectives: (i) vibe checking multimodal chat models for day-to-day tasks and (ii) rigorously testing and probing the capabilities of present frontier models. Notably, our hard set contains >50% questions that all frontier models answer incorrectly. We explore the nuances of designing, evaluating, and ranking models on ultra challenging prompts. We also discuss trade-offs between human and automatic evaluation, and show that automatic model evaluation using Reka Core roughly correlates to human judgment. We offer free API access for the purpose of lightweight evaluation and plan to conduct formal human evaluations for public models that perform well on the Vibe-Eval's automatic scores. We release the evaluation code and data, see https://github.com/reka-ai/reka-vibe-eval
@article{arxiv.2405.02287,
title = {Vibe-Eval: A hard evaluation suite for measuring progress of multimodal language models},
author = {Piotr Padlewski and Max Bain and Matthew Henderson and Zhongkai Zhu and Nishant Relan and Hai Pham and Donovan Ong and Kaloyan Aleksiev and Aitor Ormazabal and Samuel Phua and Ethan Yeo and Eugenie Lamprecht and Qi Liu and Yuqi Wang and Eric Chen and Deyu Fu and Lei Li and Che Zheng and Cyprien de Masson d'Autume and Dani Yogatama and Mikel Artetxe and Yi Tay},
journal= {arXiv preprint arXiv:2405.02287},
year = {2024}
}