Search papers, labs, and topics across Lattice.
The paper introduces M$^3$-VQA, a new VQA benchmark designed to evaluate MLLMs on fine-grained entity understanding and multi-hop reasoning across visual and textual sources. The dataset features questions requiring sequential and parallel reasoning over multiple entities, supported by a multimodal knowledge base and traceable evidence. Evaluation of 16 leading MLLMs reveals significant limitations in knowledge acquisition and reasoning, with performance improving substantially when provided with gold evidence and further enhanced by reasoning-aware retrieval methods.
Today's best multimodal LLMs still struggle to grasp fine-grained details and reason across multiple entities in images, even with access to external knowledge.
We present M$^3$-VQA, a novel knowledge-based Visual Question Answering (VQA) benchmark, to enhance the evaluation of multimodal large language models (MLLMs) in fine-grained multimodal entity understanding and complex multi-hop reasoning. Unlike existing VQA datasets that focus on coarse-grained categories and simple reasoning over single entities, M$^3$-VQA introduces diverse multi-entity questions involving multiple distinct entities from both visual and textual sources. It requires models to perform both sequential and parallel multi-hop reasoning across multiple documents, supported by traceable, detailed evidence and a curated multimodal knowledge base. We evaluate 16 leading MLLMs under three settings: without external knowledge, with gold evidence, and with retrieval-augmented input. The poor results reveal significant challenges for MLLMs in knowledge acquisition and reasoning. Models perform poorly without external information but improve markedly when provided with precise evidence. Furthermore, reasoning-aware agentic retrieval surpasses heuristic methods, highlighting the importance of structured reasoning for complex multimodal understanding. M$^3$-VQA presents a more challenging evaluation for advancing the multimodal reasoning capabilities of MLLMs. Our code and dataset are available at https://github.com/CASIA-IVA-Lab/M3VQA.
