← Back to feed

LLMs tested on historical prediction markets, NONE outperformed the market

AI · Studies · 2026-11-04 07:16

arXiv论文Forecast-Dojo构建可回放环境，用于评测和训练LLM预测智能体，在历史预测市场测试中所有模型均未跑赢市场基准。

Source: arxiv.org

Auto-generated summary for reference only. RSS