LVLM-eHub: A Comprehensive Evaluation Benchmark for Large Vision-Language Models

We are pleased to present the leaderboard for large vision-language models (LVLMs), which has been evaluated using the online platform of ⚔️ LVLM Arena. In a manner similar to CharBot Arena, LVLM Arena conducts anonymous randomized pairwise battles among LVLMs using a crowd-sourced approach. The leaderboard, based on the Elo Rating scheme, showcases the rankings of 8 LVLMs. Additionally, we provide a quantitative leaderboard across 6 categories, thoroughly assessing their multimodal capabilities.