Paper: arXiv 2502.00415
Authors: George Fatouros, Kostas Metaxas, John Soldatos, Manos Karathanassis
Abstract
MarketSenseAI is a novel framework for holistic stock analysis which leverages Large Language Models (LLMs) to process financial news, historical prices, company fundamentals and the macroeconomic environment to support decision making in stock analysis and selection. In this paper, we present the latest advancements on MarketSenseAI, driven by rapid technological expansion in LLMs. Through a novel architecture combining Retrieval-Augmented Generation and LLM agents, the framework processes SEC filings and earnings calls, while enriching macroeconomic analysis through systematic processing of diverse institutional reports. We demonstrate a significant improvement in fundamental analysis accuracy over the previous version. Empirical evaluation on S&P 100 stocks over two years (2023-2024) shows MarketSenseAI achieving cumulative returns of 125.9% compared to the index return of 73.5%, while maintaining comparable risk profiles. Further validation on S&P 500 stocks during 2024 demonstrates the framework’s scalability, delivering a 33.8% higher Sortino ratio than the market. This work marks a significant advancement in applying LLM technology to financial analysis, offering insights into the robustness of LLM-driven investment strategies.
Complexity vs Empirical Score
- Math Complexity: 2.0/10
- Empirical Rigor: 8.5/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper relies primarily on LLM agent architectures (Chain-of-Agents, RAG) rather than advanced mathematical proofs, but it demonstrates very strong empirical rigor with detailed backtests on S&P 100/500 over 2023-2024, specific performance metrics (125.9% cumulative returns vs 73.5% benchmark, 33.8% higher Sortino ratio), and real-world implementation details.
Research Flowchart
flowchart TD
A["Research Goal:<br>Enhance Stock Analysis via LLM Agents"] --> B{"Key Methodology:<br>RAG + LLM Agents"}
B --> C["Data Inputs"]
C --> D["Financial News<br>Historical Prices"]
C --> E["Company Fundamentals<br>(SEC Filings, Earnings Calls)"]
C --> F["Macroeconomic<br>(Institutional Reports)"]
D & E & F --> G["Computational Process:<br>Holistic Analysis & Selection"]
G --> H["Key Findings"]
subgraph H ["Outcomes S&P 100 2023-2024"]
H1["125.9% Cumulative Returns<br>(vs. Index 73.5%)"]
end
subgraph I ["Outcomes S&P 500 2024"]
I1["33.8% Higher Sortino Ratio<br>Comparable Risk Profile"]
end
H --> I
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 proposes the MarketSenseAI framework, featuring a Chain-of-Agents approach for granular fundamental analysis and a Retrieval-Augmented Generation module with HyDE for macroeconomic context.
Source sentence
We introduce a Chain-of-Agents (CoA) approach that enables granular handling of large-scale financial data—such as 10-Q, 10-K
Assumes 2
- Transaction costs are 10 basis points per trade and are deducted from reported returns. execution
Source sentence
Values in parentheses represent the total returns (%) after transaction costs (10bps/trade).
- Portfolios are rebalanced monthly based on the buy signals generated by MarketSenseAI. execution
Source sentence
Equally weighted portfolio rebalanced monthly based on the buy signals of MarketSenseAI
Data
2023 – 2024 · S&P 100 stocks · equities · monthly
- S&P 100 stock universe benchmark public
Source sentence
Empirical evaluation on S&P 100 stocks over two years (2023-2024) shows MarketSenseAI achieving cumulative returns of 125.9% compared to the index return of 73.5%
- S&P 500 stock universe benchmark public
Source sentence
Further validation on S&P 500 stocks during 2024 demonstrates the framework's scalability, delivering a 33.8% higher Sortino ratio than the market.
SEC EDGAR
Costs
Modeled
10 basis points per trade are subtracted from returns (values in parentheses).
Source sentence
Values in parentheses represent the total returns (%) after transaction costs (10bps/trade).
Finds 4
Sharpe 2.76 (benchmark 2.52) · total return 125.9% · max drawdown 13.8%
- MarketSenseAI achieved a 125.9% cumulative return on the S&P 100 cap‑weighted portfolio, outperforming the index's 73.5% return. performance
Source sentence
In the S&P 100 universe, the system’s selected stocks achieved a 125.9% cumulative return under market cap‑weighting (MS‑Cap), significantly surpassing the S&P 100 index return of 73.5% (Table V).
- The MS‑Cap portfolio had a Sortino ratio 16% higher than the benchmark (4.43 vs 3.82) despite higher volatility. performance
Source sentence
the MS‑Cap portfolio exhibited a 16% higher Sortino ratio (4.43 vs. 3.82) compared to the cap‑weighted benchmark, despite experiencing higher volatility.
- On the S&P 500 in 2024, MarketSenseAI’s equal‑weighted portfolio returned 25.8% versus 12.8% for the benchmark, a 102% relative outperformance. performance
Source sentence
returns in equal‑weighted portfolios compared to 12.8% for the S&P 500 Equal Weight benchmark, representing a 102% relative outperformance.
- The framework’s macroeconomic retrieval methods achieved high context precision (≥0.98) across configurations. performance
Source sentence
Context Precision remains high (≥0.98) in all configurations, indicating that even when queries span multiple reports, irrelevant chunks are not in the top‑n results.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 38 related papers). Coverage is partial, so this is neither replication nor refutation.
5 papers in the archive
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.