![[2508.21285] A Financial Brain Scan of the LLM](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2508.21285] A Financial Brain Scan of the LLM
Summary
This article presents a novel approach to analyzing large language models (LLMs) in finance, enabling researchers to identify and manipulate the underlying concepts that influence LLM-generated economic forecasts.
Why It Matters
Understanding how LLMs generate economic forecasts is crucial for researchers and practitioners in finance. This method enhances transparency and allows for the correction of biases in model predictions, making it a valuable tool for empirical research in social sciences.
Key Takeaways
- The study introduces a method to 'brain scan' LLMs, revealing the concepts guiding their reasoning.
- Researchers can steer LLMs to adjust their risk profiles and biases, enhancing prediction accuracy.
- The approach is transparent, lightweight, and replicable, facilitating empirical research.
Quantitative Finance > General Finance arXiv:2508.21285 (q-fin) [Submitted on 29 Aug 2025 (v1), last revised 13 Feb 2026 (this version, v2)] Title:A Financial Brain Scan of the LLM Authors:Hui Chen, Antoine Didisheim, Mohammad (Mo)Pourmohammadi, Luciano Somoza, Hanqing Tian View a PDF of the paper titled A Financial Brain Scan of the LLM, by Hui Chen and 4 other authors View PDF HTML (experimental) Abstract:Emerging techniques in computer science make it possible to "brain scan" large language models (LLMs), identify the plain-English concepts that guide their reasoning, and steer them while holding other factors constant. We show that this approach can map LLM-generated economic forecasts to concepts such as sentiment, technical analysis, and timing, and compute their relative importance without reducing performance. We also show that models can be steered to be more or less risk-averse, optimistic, or pessimistic, which allows researchers to correct or simulate biases. The method is transparent, lightweight, and replicable for empirical research in the social sciences. Comments: Subjects: General Finance (q-fin.GN); Artificial Intelligence (cs.AI); Computational Engineering, Finance, and Science (cs.CE); General Economics (econ.GN) Cite as: arXiv:2508.21285 [q-fin.GN] (or arXiv:2508.21285v2 [q-fin.GN] for this version) https://doi.org/10.48550/arXiv.2508.21285 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Hanqing Tian [view email] [v1] Fr...
