DeepFinance 基准：AI 能否尽调一亿美元级交易？
DeepFinance: Can AI Diligence a $100MM Deal?
arXiv 上这个新基准把 21 项真实 PE/投行任务交给 22 个模型跑，GPT-6 Astra 也只满足 57.6% 的评分标准。想看 AI 在尽调这种长链路、要审计的高风险工作里到底几斤几两的，值得把论文里的失败案例翻一遍。
DeepFinance 是一个面向私募股权（PE）与投行（IB）真实业务的超长程基准，包含 21 项端到端任务，人类专业人员平均完成一项需 27.1 小时。每项任务配有平均 19 份文件的数据室，并按平均 84 条专家撰写的评分细则打分，覆盖输出质量、工作流程与商业判断三个维度。
研究团队用各模型原生的 agent 框架评测了 22 个模型配置。成绩最好的是开启最大推理的 GPT-6 Astra，满足 57.6% 的评分细则，且排名前四的配置在统计上无法区分。评测揭示了两类典型失败：模型会用错误的输入算出正确的数字，也能找到影响决策的关键证据却不据此行动。
与以往聚焦文档检索和数值正确性的金融评测不同，DeepFinance 考察的是在相互冲突的信息源之间做取舍、设立合理假设并保持全流程可审计的能力，这正是 PE 与投行工作的核心。
DeepFinance: Can AI Diligence a $100MM Deal?
Existing evaluations of AI models on high-finance work focus on retrieving answers from financial documents and producing numerically correct outputs. Real private equity (PE) and investment banking (IB) work also depends on choosing between conflicting sources, making sound assumptions, and keeping every step auditable. We introduce DeepFinance, an ultra long-horizon benchmark of 21 end-to-end PE and IB tasks that take a human professional an average of 27.1 hours. Each task pairs a request with a data room averaging 19 files and is graded against an average of 84 expert-written rubric criteria covering output, workflow, and judgment. We evaluate 22 model configurations, each in its native agent harness. The best configuration, GPT-6 Astra with max reasoning, satisfies 57.6% of rubric criteria, and the top four configurations are not statistically distinguishable. Frontier models often calculate correctly from the wrong inputs and find decision-relevant evidence without acting on it.
This report is categorized under AITOP's "Papers" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include DeepFinance, GPT-6, 私募股权, 投行.
Source attribution: curated from arXiv cs.AI (@Drew Hahn, Tyler Shibata, Jackson Grove, Rahul Sonwalkar). Cross-source corroboration: 10 independent sources have covered this development (including 宝玉, AIGCLINK, Yuchen Jin), confirming industry-wide signal strength.