VisNumBench: Evaluating Number Sense of Multimodal Large Language Models
This provides a new benchmark for the research community to assess and improve number sense in MLLMs, but it is incremental as it focuses on evaluation rather than novel solutions.
The authors tackled the problem of evaluating whether multimodal large language models (MLLMs) can develop human-like intuitive number sense by introducing VisNumBench, a benchmark with about 1,900 multiple-choice questions across visual numerical tasks, and found that 17 tested MLLMs performed significantly below human levels, with only modest gains from larger models.
Can Multimodal Large Language Models (MLLMs) develop an intuitive number sense similar to humans? Targeting this problem, we introduce Visual Number Benchmark (VisNumBench) to evaluate the number sense abilities of MLLMs across a wide range of visual numerical tasks. VisNumBench consists of about 1,900 multiple-choice question-answer pairs derived from both synthetic and real-world visual data, covering seven visual numerical attributes and four types of visual numerical estimation tasks. Our experiments on VisNumBench led to the following key findings: (i) The 17 MLLMs we tested, including open-source models such as Qwen2.5-VL and InternVL2.5, as well as proprietary models like GPT-4o and Gemini 2.0 Flash, perform significantly below human levels in number sense-related tasks. (ii) Multimodal mathematical models and multimodal chain-of-thought (CoT) models did not exhibit significant improvements in number sense abilities. (iii) Stronger MLLMs with larger parameter sizes and broader general abilities demonstrate modest gains in number sense abilities. We believe VisNumBench will serve as a valuable resource for the research community, encouraging further advancements in enhancing MLLMs' number sense abilities. Code and dataset are available at https://wwwtttjjj.github.io/VisNumBench/.