Paper: arXiv 2410.07225
Authors: Chung-Chi Chen, Hiroya Takamura, Ichiro Kobayashi, Yusuke Miyao
Abstract
Professionals’ decisions are the focus of every field. For example, politicians’ decisions will influence the future of the country, and stock analysts’ decisions will impact the market. Recognizing the influential role of professionals’ perspectives, inclinations, and actions in shaping decision-making processes and future trends across multiple fields, we propose three tasks for modeling these decisions in the financial market. To facilitate this, we introduce a novel dataset, A3, designed to simulate professionals’ decision-making processes. While we find current models present challenges in forecasting professionals’ behaviors, particularly in making trading decisions, the proposed Chain-of-Decision approach demonstrates promising improvements. It integrates an opinion-generator-in-the-loop to provide subjective analysis based on each news item, further enhancing the proposed tasks’ performance.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 7.0/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper introduces a novel dataset with specific train/test splits and task metrics, demonstrating strong empirical implementation, but relies on standard NLP/ML models (LLM prompting, classification) without heavy mathematical derivations or advanced stochastic calculus.
Research Flowchart
flowchart TD
A["Research Goal:<br>Modeling Professionals'<br>Decisions in Finance"] --> B{"Key Methodology"}
B --> C["Dataset: A3<br>Simulates Decision Processes"]
B --> D["Method: Chain-of-Decision<br>with Opinion-Generator"]
C --> E["Computational Process:<br>News & Opinion Integration"]
D --> E
E --> F["Key Findings:<br>1. Current models struggle<br>2. Chain-of-Decision improves<br>performance"]
F --> G["Outcome:<br>Enhanced Financial<br>Forecasting & Analysis"]
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 three new tasks for modeling professional analysts' decisions and a Chain-of-Decision approach that incorporates an opinion-generator-in-the-loop.
Source sentence
we propose three tasks to evaluate whether the proposed Chain-of-Decision approach enables models to make decisions like professionals: (1) deciding whether to share the formed opinion with clients, (2) determining if the news alters their initial perspective on the company, and (3) choosing whether to buy or sell the
Assumes 3
- Professionals’ decisions are predominantly shaped by the most recent news. model
Source sentence
our approach assumes that professionals’ decisions are predominantly shaped by the most recent news.
- Real‑time, accurate news data is available for the model to predict behavior. data
Source sentence
Real-time access to accurate and comprehensive news data is necessary for the model to predict the behaviors of professionals effectively.
- The quality and accuracy of the generated opinions affect model performance. model
Source sentence
our approach relies on the quality and accuracy of the generated opinions.
Data
Costs
Not applicable
Finds 3
- The Chain-of-Decision approach yields promising improvements in task performance. performance
Source sentence
Our experiments showed promising improvements in the performance of the tasks at hand when employing the CoD approach.
- mT5 underperforms on generated analysis but provides the most informative inputs for decision‑making models. performance
Source sentence
although mT5 underperforms when assessed on the generated analysis, it provides the most informative inputs for decision-making models.
- Using DAN‑guided ChatGPT improves performance compared to the standard version. performance
Source sentence
we observe enhanced performance when utilizing the DAN-guided ChatGPT, as compared to its standard version.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (36)
- Firm-specific or market-wide information: How does analyst coverage influence stock price synchronicity? (2022)
- SentiMI: Introducing point-wise mutual information with SentiWordNet to improve sentiment polarity detection (2016)
- CPT : A pre-trained unbalanced transformer for both chinese language understanding and generation (2021) — link
- Modeling Financial Analysts ’ Decision Making via the Pragmatics and Semantics of Earnings Calls (2019)
- How do analyst recommendations respond to major news? (2006)
- Investor reactions to financial analysts’ research reports (1995)
- From opinion mining to financial argument mining (2021)
- Decision-Focused Learning: Through the Lens of Learning to Rank (2022)
and 28 more
- UER: An Open-Source Toolkit for Pre-training Models (2019)
- Attention is all you need (2017)
- XL -Sum: Large-Scale Multilingual Abstractive Summarization for 44 Languages (2021)
- Transformers: State-of-the-Art Natural Language Processing (2020)
- BERT : Pre-training of Deep Bidirectional Transformers for Language Understanding (2019)
- Pegasus: Pre-training with extracted gap-sentences for abstractive summarization (2020)
- Mengzi: Towards Lightweight yet Ingenious Pre-trained Models for Chinese (2021)
- m T 5: A Massively Multilingual Pre-trained Text-to-Text Transformer (2021)
- L ink BERT : Pretraining Language Models with Document Links (2022)
- Melding the data-decisions pipeline: Decision-focused learning for combinatorial optimization (2019)
- Learning to Generate Market Comments from Stock Prices (2017)
- Market Comment Generation from Data with Noisy Alignments (2020)
- An Empirical Investigation of Bias in the Multimodal Analysis of Financial Earnings Calls (2021)
- Buy Tesla, Sell Ford: Assessing Implicit Stock Market Preference in Pre-trained Language Models (2022)
- The impact of sell-side analyst research coverage on an affiliated broker’s market share of trading volume (2010)
- Analysts’ forecasts and asset pricing (2016)
- Market liquidity and trading activity (2001)
- Trading activity and expected stock returns (2001)
- Economic uncertainty and national bitcoin trading activity (2021)
- Stock Movement Prediction from Tweets and Historical Prices (2018)
- Maec: A multimodal aligned earnings conference call dataset for financial risk prediction (2020)
- What You Say and How You Say It Matters: Predicting Stock Volatility Using Verbal and Vocal Cues (2019)
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models (2022)
- Modeling the Stock Relation with Graph Network for Overnight Stock Movement Prediction. (2020)
- Random walks in stock market prices (1995)
- Analyst report readability and stock returns (2016)
- The Pricing of Options and Corporate Liabilities (1973)
- Stock return synchronicity and the market response to analyst recommendation revisions (2015)
0 of 36 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.