OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problems

TL;DR

OlympiadBench is a challenging bilingual multimodal benchmark with 8,476 Olympiad-level problems designed to evaluate and advance the reasoning abilities of large AI models, especially in scientific domains like mathematics and physics.

Contribution

This work introduces OlympiadBench, a new rigorous benchmark with expert annotations, to better assess AI models' advanced reasoning and problem-solving skills in scientific competitions.

Findings

GPT-4V scores 17.97% on OlympiadBench

Physics problems are particularly challenging with 10.74% average score

Common issues include hallucinations, knowledge omissions, and logical fallacies

Abstract

Recent advancements have seen Large Language Models (LLMs) and Large Multimodal Models (LMMs) surpassing general human capabilities in various tasks, approaching the proficiency level of human experts across multiple domains. With traditional benchmarks becoming less challenging for these models, new rigorous challenges are essential to gauge their advanced abilities. In this work, we present OlympiadBench, an Olympiad-level bilingual multimodal scientific benchmark, featuring 8,476 problems from Olympiad-level mathematics and physics competitions, including the Chinese college entrance exam. Each problem is detailed with expert-level annotations for step-by-step reasoning. Evaluating top-tier models on OlympiadBench, we implement a comprehensive assessment methodology to accurately evaluate model responses. Notably, the best-performing model, GPT-4V, attains an average score of 17.97% on OlympiadBench, with a mere 10.74% in physics, highlighting the benchmark rigor and the intricacy of physical reasoning. Our analysis orienting GPT-4V points out prevalent issues with hallucinations, knowledge omissions, and logical fallacies. We hope that our challenging benchmark can serve as a valuable resource for helping future AGI research endeavors. The data and evaluation code are available at \url{https://github.com/OpenBMB/OlympiadBench}