AGIBench: A Multi-granularity, Multimodal, Human-Referenced, Auto-Scoring Benchmark for Large Language Models

Large language models (LLMs) like ChatGPT have revealed amazing intelligence. How to evaluate the question-solving abilities of LLMs and their degrees of intelligence is a hot-spot but challenging issue. First, the question-solving abilities are interlaced with different ability branches like understanding and massive knowledge categories like mathematics. Second, the inputs of questions are multimodal that may involve text and images. In addition, they may have varying levels of difficulty while lacking a unified standard to judge which one is more difficult. Third, the response format of LLMs is diverse and thus poses great challenges for result extraction and evaluation. Several benchmarks have been proposed to evaluate LLMs, yet they still exhibit significant shortcomings. In this paper, to tackle the above challenges, we propose AGIBench—a multi-granularity, multimodal, human-referenced, and auto-scoring benchmarking methodology for LLMs. Instead of a collection of blended questions, AGIBench focuses on three typical ability branches and adopts a four-tuple to label the attributes of each question. First, it supports multi-granularity benchmarking. Second, it contains multimodal input, including text and images. Third, it classifies all the questions into five degrees of difficulty according to the average accuracy rate of abundant educated humans (human-referenced). Fourth, it adopts zero-shot learning to avoid introducing additional unpredictability and provides an auto-scoring method to extract and judge the result. Finally, it defines multi-dimensional metrics. Our experiments on twelve state-of-the-art LLMs show the effectiveness of our benchmark. AGIBench is publicly available from https://www.benchcouncil.org/agibench .