Crypto AI Agent Benchmark (CAIA)
收藏资源简介:
CAIA是一个基准测试,旨在评估AI代理在对抗性、高风险环境下的能力,特别是在加密货币市场中,代理需要从操纵中辨别真相,导航碎片化的信息景观,并在对抗性条件下做出不可逆转的金融决策。该数据集包含178个基于时间的任务,用于测试代理的推理能力、规划能力和行动能力。数据集来源于超过10,000个真实的查询,通过一个五阶段的严格筛选流程,保证了任务的质量和真实性。
CAIA is a benchmark designed to evaluate the capabilities of AI Agents in adversarial, high-stakes environments, particularly within cryptocurrency markets, where agents must distinguish truth from manipulation, navigate fragmented information landscapes, and make irreversible financial decisions under adversarial conditions. This dataset includes 178 time-based tasks to test agents' reasoning, planning, and operational abilities. Derived from over 10,000 real-world queries, the dataset utilizes a rigorous five-stage screening process to ensure the quality and authenticity of the tasks.

- 1通过Surf AI, Cybertino Lab, Princeton University · 2025年



