Position: Evaluating LLMs in Finance Requires Explicit Bias Consideration

Yaxuan Kong, Hoyoung Lee, Yoontae Hwang, Alejandro Lopez-Lira, Bradford Levy, Dhagash Mehta, Qingsong Wen, Chanyeol Choi, Yongjae Lee, Stefan Zohren
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:171089-171110, 2026.

Abstract

Large Language Models (LLMs) are increasingly integrated into financial workflows, but evaluation practice has not kept up. Finance-specific biases can inflate performance, contaminate backtests, and make reported results useless for any deployment claim. We identify five recurring biases in financial LLM applications. They include look-ahead bias, survivorship bias, narrative bias, objective bias, and cost bias. These biases break financial tasks in distinct ways and they often compound to create an illusion of validity. We reviewed 164 papers from 2023 to 2025 and found that no single bias is discussed in more than 28 percent of studies. This position paper argues that bias in financial LLM systems requires explicit attention and that structural validity should be enforced before any result is used to support a deployment claim. We propose a Structural Validity Framework and an evaluation checklist with minimal requirements for bias diagnosis and future system design. The material is available at https://github.com/Eleanorkong/Awesome-Financial-LLM-Bias-Mitigation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-kong26i, title = {Position: Evaluating {LLM}s in Finance Requires Explicit Bias Consideration}, author = {Kong, Yaxuan and Lee, Hoyoung and Hwang, Yoontae and Lopez-Lira, Alejandro and Levy, Bradford and Mehta, Dhagash and Wen, Qingsong and Choi, Chanyeol and Lee, Yongjae and Zohren, Stefan}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {171089--171110}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/kong26i/kong26i.pdf}, url = {https://proceedings.mlr.press/v306/kong26i.html}, abstract = {Large Language Models (LLMs) are increasingly integrated into financial workflows, but evaluation practice has not kept up. Finance-specific biases can inflate performance, contaminate backtests, and make reported results useless for any deployment claim. We identify five recurring biases in financial LLM applications. They include look-ahead bias, survivorship bias, narrative bias, objective bias, and cost bias. These biases break financial tasks in distinct ways and they often compound to create an illusion of validity. We reviewed 164 papers from 2023 to 2025 and found that no single bias is discussed in more than 28 percent of studies. This position paper argues that bias in financial LLM systems requires explicit attention and that structural validity should be enforced before any result is used to support a deployment claim. We propose a Structural Validity Framework and an evaluation checklist with minimal requirements for bias diagnosis and future system design. The material is available at https://github.com/Eleanorkong/Awesome-Financial-LLM-Bias-Mitigation.} }
Endnote
%0 Conference Paper %T Position: Evaluating LLMs in Finance Requires Explicit Bias Consideration %A Yaxuan Kong %A Hoyoung Lee %A Yoontae Hwang %A Alejandro Lopez-Lira %A Bradford Levy %A Dhagash Mehta %A Qingsong Wen %A Chanyeol Choi %A Yongjae Lee %A Stefan Zohren %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-kong26i %I PMLR %P 171089--171110 %U https://proceedings.mlr.press/v306/kong26i.html %V 306 %X Large Language Models (LLMs) are increasingly integrated into financial workflows, but evaluation practice has not kept up. Finance-specific biases can inflate performance, contaminate backtests, and make reported results useless for any deployment claim. We identify five recurring biases in financial LLM applications. They include look-ahead bias, survivorship bias, narrative bias, objective bias, and cost bias. These biases break financial tasks in distinct ways and they often compound to create an illusion of validity. We reviewed 164 papers from 2023 to 2025 and found that no single bias is discussed in more than 28 percent of studies. This position paper argues that bias in financial LLM systems requires explicit attention and that structural validity should be enforced before any result is used to support a deployment claim. We propose a Structural Validity Framework and an evaluation checklist with minimal requirements for bias diagnosis and future system design. The material is available at https://github.com/Eleanorkong/Awesome-Financial-LLM-Bias-Mitigation.
APA
Kong, Y., Lee, H., Hwang, Y., Lopez-Lira, A., Levy, B., Mehta, D., Wen, Q., Choi, C., Lee, Y. & Zohren, S.. (2026). Position: Evaluating LLMs in Finance Requires Explicit Bias Consideration. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:171089-171110 Available from https://proceedings.mlr.press/v306/kong26i.html.

Related Material