arXiv cs.AI· Saanvi Khetan, Sankar Balasubramanian·· 10 天前AI 评分32
Llama-3.1-8B-Instruct 研究：财务信息缺失时，语言模型如何用身份替代事实
Thin Evidence, Thick Priors: How Language Models Substitute Identity for Missing Financial Facts
AI 导读
一项针对 Llama-3.1-8B-Instruct 的 96,600 条提示测试发现，在财务信息完全相同的情况下，仅改变投资者身份，推荐股票配置的平均差距会从信息完整披露时的 4.78 个百分点升至无任何财务事实时的 10.34 个百分点，比值达 2.16。
正文 · 原文
Abstract:People increasingly ask large language models what to do with their money, yet seldom describe their finances in full. This paper asks what a model does with the gap. Holding finances fixed and changing only who the investor is said to be, we grade the financial evidence in the prompt from eight facts to none and measure how far the recommended equity allocation moves. Across 96,600 prompts to Llama-3.1-8B-Instruct, built from 100 financial profiles, 138 personas and seven disclosure conditions, the average gap between two personas with identical finances rises from 4.78 percentage points at full disclosure to 10.34 points with no financial facts. A two-way cluster bootstrap counting duplicated prompts once places the ratio at 2.16 (95% interval 1.69 to 2.79), and the rise is already 1.69-fold with a single fact left. Identity explains 5% of within-profile variation in advice at full disclosure and 96% with no disclosure. Household size is the only attribute whose influence grows reliably as evidence is withdrawn. Once standard errors are clustered on the persona, the unit to which identity was assigned, most attribute-specific interactions reported in the conference version lose significance, and gender instead appears as a small standing gap that full disclosure does not close. Stating risk appetite alone brings the swing into the range seen with two to seven generic facts. With no facts, the model's one-line rationale cites incomes, debts and savings it was never told, and these invented finances turn adverse more often for larger households. Inside the network, gender is linearly decodable at every layer, and ablating the gender direction at five layers leaves the aggregate identity swing unchanged. Advisory systems built on such models should be audited at the disclosure levels users actually reach, and judged across the whole identity space rather than one attribute at a time.
| Comments: | 49 pages, 17 figures, 12 tables; Submitted & Accepted to ICAIF'2026 |
| Subjects: | Artificial Intelligence (cs.AI) |
| MSC classes: | 68T50, 91G80, 62P20 |
| ACM classes: | J.4; I.2.7; K.4.2 |
| Cite as: | arXiv:2610.07798 [cs.AI] |
| (or arXiv:2610.07798v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2610.07798 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Sankar B [view email]
[v1]
Tue, 6 Oct 2026 05:46:28 UTC (2,484 KB)
来源：arXiv cs.AI · arxiv.org