Paper: arXiv 2505.17048

Authors: Agam Shah, Siddhant Sukhani, Huzaifa Pardawala, Saketh Budideti, Riya Bhadani, Rudra Gopal, Siddhartha Somani, Rutwik Routu, Michael Galarnyk, Soungmin Lee, Arnav Hiray, Akshar Ravichandran, Eric Kim, Pranav Aluru, Joshua Zhang, Sebastian Jaskowski, Veer Guda, Meghaj Tarte, Liqin Ye, Spencer Gosden, Rachel Yuh, Sloka Chava, Sahasra Chava, Dylan Patrick Kelly, Aiden Chiang, Harsit Mittal, Sudheer Chava

Abstract

Central banks around the world play a crucial role in maintaining economic stability. Deciphering policy implications in their communications is essential, especially as misinterpretations can disproportionately impact vulnerable populations. To address this, we introduce the World Central Banks (WCB) dataset, the most comprehensive monetary policy corpus to date, comprising over 380k sentences from 25 central banks across diverse geographic regions, spanning 28 years of historical data. After uniformly sampling 1k sentences per bank (25k total) across all available years, we annotate and review each sentence using dual annotators, disagreement resolutions, and secondary expert reviews. We define three tasks: Stance Detection, Temporal Classification, and Uncertainty Estimation, with each sentence annotated for all three. We benchmark seven Pretrained Language Models (PLMs) and nine Large Language Models (LLMs) (Zero-Shot, Few-Shot, and with annotation guide) on these tasks, running 15,075 benchmarking experiments. We find that a model trained on aggregated data across banks significantly surpasses a model trained on an individual bank’s data, confirming the principle “the whole is greater than the sum of its parts.” Additionally, rigorous human evaluations, error analyses, and predictive tasks validate our framework’s economic utility. Our artifacts are accessible through the HuggingFace and GitHub under the CC-BY-NC-SA 4.0 license.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 9.0/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper’s mathematics is relatively low, primarily relying on standard statistical measures and NLP metrics without complex derivations, while its empirical rigor is extremely high, featuring a large curated dataset (25k annotated sentences), extensive benchmarking (15,075 experiments), and explicit release of models and code for reproducibility.

Research Flowchart

  flowchart TD
  A["Research Goal<br>Decipher Central Bank<br>Communications Globally"] --> B["Data Collection<br>WCB Dataset: 380k sentences<br>25 banks, 28 years"]
  B --> C["Annotation & Labeling<br>25k sampled sentences<br>Stance, Temporal, Uncertainty"]
  C --> D["Computational Benchmarking<br>7 PLMs + 9 LLMs<br>15k+ experiments"]
  D --> E{"Key Findings & Outcomes"}
  E --> F["Model Performance<br>Training on aggregated data<br>surpasses individual banks"]
  E --> G["Framework Validation<br>Human eval & error analysis<br>confirm economic utility"]
  E --> H["Artifacts Released<br>HuggingFace & GitHub<br>CC-BY-NC-SA 4.0"]

Evidence Card

What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card

Proposes

The paper introduces the World Central Banks (WCB) dataset, a large-scale monetary policy corpus, and releases 78 fine‑tuned pretrained language models for extracting insights from central‑bank communications.

Source sentence
we introduce the World Central Banks (WCB) dataset, the most comprehensive monetary policy corpus to date, comprising over 380k sentences from 25 central banks across diverse geographic regions, spanning 28 years of historical data.

Assumes

No explicit assumptions found in the text.

Data

  • World Central Banks (WCB) dataset text-news public
    1996-2024 · 25 central banks
    Source sentence
    The World Central Banks (WCB) dataset is distributed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC-BY-NC-SA 4.0) license

Recognised sources: Bloomberg Kaggle

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • A model trained on aggregated data across banks significantly surpasses a model trained on an individual bank's data. performance
    Source sentence
    We find that a model trained on aggregated data across banks significantly surpasses a model trained on an individual bank's data, confirming the principle "the whole is greater than the sum of its parts."
  • Training on the whole corpus yields significant improvement in performance across all three tasks compared to bank‑specific training. performance
    Source sentence
    In our experiments, we discover that there is a significant improvement in model performance across all three labels when the training corpus of the is expanded from just the specific bank (Bank Specific Setup) to having all banks’ data (General Setup).
  • Few‑shot and annotation‑guide prompting give minor gains for Stance Detection but LLMs still underperform fine‑tuned PLMs. performance
    Source sentence
    however, even with enhanced prompting, LLMs do not surpass PLM performance, suggesting that the performance ceiling for LLMs remains lower than that of fine‑tuned PLMs on the Stance Detection task.

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 15 related papers). Coverage is partial, so this is neither replication nor refutation.

Code

Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.