Unverified75% confidenceFactTime unknown
2023年人类在GAIA测试中的平均成功率约为90%，而当时最强的GPT-4在Level 1上仅勉强达到15%
1
Sources
75%
Confidence
Long-term
Relevance
6/1/2026
First Seen
Sources
Related Entities
Related Claims
Unverified自2023年GPT-4发布以来，AI在HumanEval等主流编程基准测试上的通过率从不足50%跃升至90%以上72% similarUnverifiedThe pass rate of top models in 2025 on the HumanEval benchmark has exceeded 90%.69% similarUnverifiedARC-AGI-2由AI安全研究员François Chollet设计，早期GPT-4在该测试上得分不足10%，人类平均得分约85%68% similarUnverifiedCurrent top commercial models typically score no more than 10% on GAIA Level 3.68% similarUnverifiedGAIA题目对人类来说相对简单，人类平均正确率约92%，但对顶级AI模型极具挑战性66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/20910API
curl https://kongchang.com/api/v1/knowledge/claims/20910MCP
get_claim(id=20910)