MIBench

MIBench is a benchmark designed to evaluate the fine-grained abilities of multimodal large language models (MLLMs) in multi-image scenarios, comprising 13 tasks with a total of 13,000 annotated samples across three categories: multi-image instruction, multimodal knowledge-seeking, and multimodal in-context learning.

Papers using MIBench (5)

OMIBench: Benchmarking Olympiad-Level Multi-Image Reasoning in Large Vision-Language Model2026FlexID: Training-Free Flexible Identity Injection via Intent-Aware Modulation for Text-to-Image Generation2026MIHBench: Benchmarking and Mitigating Multi-Image Hallucinations in Multimodal Large Language Models2025Closing the Modality Gap for Mixed Modality Search2025MIBench: Evaluating Multimodal Large Language Models over Multiple Images2024 · 1 cites