Modeling Hawkish-Dovish Latent Beliefs in Multi-Agent Debate-Based LLMs for Monetary Policy Decision Classification
Kaito Takano, Masanori Hirano, Kei Nakagawa · 2025
Read Modeling Hawkish-Dovish Latent Beliefs in Multi-Agent Debate-Based LLMs for Monetary Policy Decision Classification by Kaito Takano, Masanori Hirano, Kei Nakagawa (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-02469.txt.
Description
Kaito Takano, Masanori Hirano, and Kei Nakagawa forecast central bank decisions by simulating the committee that makes them. Their subject is the Federal Open Market Committee, whose rate decisions are usually predicted from policy texts with static classifiers that ignore the deliberation behind the decision. The authors instead field a panel of language-model agents, each starting from its own initial belief, reading the same qualitative texts and macroeconomic indicators.
Over several rounds the agents see each other's predictions and revise their own, mimicking deliberation and consensus formation. A latent variable tracks each agent's underlying stance, hawkish or dovish. The theory shows how that stance shapes both what an agent notices in the data and how it responds to peers. The debate-based system significantly beat standard language-model baselines on prediction accuracy, and the explicit beliefs made the forecasts easier to interpret.
Catalog record
| CREATOR | Kaito Takano, Masanori Hirano, Kei Nakagawa |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-02469