Investor's Last Exam evaluates AI agents on equity research
An arXiv paper submitted on October 28, 2026 introduces Investor's Last Exam (ILE), an equity-research benchmark led by experienced buy-side investors and fund managers. ILE uses a versioned financial database with 3.43 million document versions, validated decision-time information, professional analyst tasks, and expert-led scoring across 86 assignments. It jointly assesses research quality, forecast accuracy, and simulated investment performance. The paper reports that GPT-6 Astra and Claude Opus 5.5 led overall with scores of 68.1 and 66.9 out of 100, while Astra received 52.5 on Forecast & Model.
ILE evaluates the evidence-assumption-judgment chain rather than isolated abilities, enabling researchers to trace where strong overall research performance fails to translate into forecasts and investment judgments.
Computer Science > Artificial Intelligence
arXiv:2610.38342 (cs)
[Submitted on 28 Oct 2026]
Title:Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?
Authors:Huangfei Jiang, Anning Hu, Kai Sheng, Yuchen Cai, Rui Ge, Xiangyu Kong, Jianghao Lin, Shuzi Chen, Jiyi Weng, Junchi Yan
View a PDF of the paper titled Investor's Last Exam: Can AI Agents Do Equity Investment Without Seeing the Future?, by Huangfei Jiang and 9 other authors