yujunjun/TradingAgents
1
0
Fork 0
mirror of https://github.com/TauricResearch/TradingAgents.git synced 2025-06-27 00:41:34 +00:00
Code Issues Packages Projects Releases Wiki Activity

TradingAgents-AI

Browse Source
This commit is contained in:
Yijia-Xiao 2024-12-28 11:32:06 +08:00
commit c2fa046a9b
2 changed files with 460 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
README.md Normal file
View File

@ -0,0 +1,5 @@
# TradingAgents: Multi-Agents LLM Financial Trading Framework
> MARW Workshop, AAAI 2025
>
> Homepage: https://TradingAgents.github.io/

455
index.html Normal file
View File

@ -0,0 +1,455 @@
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="description" content="TradingAgents: Multi-Agents LLM Financial Trading Framework">
<meta name="keywords" content="TradingAgents, LLM, Financial Trading, Multi-Agent Systems, Financial Markets, AI Trading">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>TradingAgents: Multi-Agents LLM Financial Trading Framework</title>
<script async src="https://www.googletagmanager.com/gtag/js?id=G-PYVRSFMDRL"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() { dataLayer.push(arguments); }
gtag('js', new Date());
gtag('config', 'G-PYVRSFMDRL');
</script>
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<link rel="stylesheet" href="./static/css/bulma.min.css">
<link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./static/css/bulma-slider.min.css">
<link rel="stylesheet" href="./static/css/fontawesome.all.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="icon" href="./static/images/TradingAgents.png">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script defer src="./static/js/fontawesome.all.min.js"></script>
<script src="./static/js/bulma-carousel.min.js"></script>
<script src="./static/js/bulma-slider.min.js"></script>
<script src="./static/js/index.js"></script>
</head>
<body>
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<a class="navbar-item" href="https://yijia-xiao.github.io/">
<span class="icon">
<i class="fas fa-home"></i>
</span>
</a>
<div class="navbar-item has-dropdown is-hoverable">
<a class="navbar-link">More Research</a>
<div class="navbar-dropdown">
<a class="navbar-item" href="https://arxiv.org/abs/2408.11363">ProteinGPT</a>
<a class="navbar-item" href="https://arxiv.org/abs/2310.02469">PrivacyMind</a>
<a class="navbar-item" href="https://arxiv.org/abs/XXXX.XXXXX">TradingAgents</a>
</div>
</div>
</div>
</div>
</nav>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title">TradingAgents: Multi-Agents LLM Financial Trading Framework</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">Yijia Xiao<sup>1</sup>,</span>
<span class="author-block">Edward Sun<sup>1</sup>,</span>
<span class="author-block">Di Luo<sup>2</sup>,</span>
<span class="author-block">Wei Wang<sup>1</sup></span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup>1</sup>University of California, Los Angeles,</span>
<span class="author-block"><sup>2</sup>Massachusetts Institute of Technology</span>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<span class="link-block"><a href="https://arxiv.org/abs/XXXX.XXXXX" class="external-link button is-normal is-rounded is-dark"><span class="icon"><i class="fas fa-file-pdf"></i></span><span>Paper</span></a></span>
<span class="link-block"><a href="https://github.com/Yijia-Xiao/TradingAgents" class="external-link button is-normal is-rounded is-dark"><span class="icon"><i class="fab fa-github"></i></span><span>Code</span></a></span>
<!-- Add more links if available -->
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Abstract</h2>
<div class="content has-text-justified">
<p>Significant progress has been made in automated problem-solving using societies of agents powered by large language models (LLMs). In finance, efforts have largely focused on single-agent systems handling specific tasks or multi-agent frameworks independently gathering data. However, multi-agent systems' potential to replicate real-world trading firms' collaborative dynamics remains underexplored. <strong>TradingAgents</strong> proposes a novel stock trading framework inspired by trading firms, featuring LLM-powered agents in specialized roles such as fundamental analysts, sentiment analysts, technical analysts, and traders with varied risk profiles. The framework includes Bull and Bear researcher agents assessing market conditions, a risk management team monitoring exposure, and traders synthesizing insights from debates and historical data to make informed decisions. By simulating a dynamic, collaborative trading environment, this framework aims to improve trading performance. Detailed architecture and extensive experiments reveal its superiority over baseline models, with notable improvements in cumulative returns, Sharpe ratio, and maximum drawdown, highlighting the potential of multi-agent LLM frameworks in financial trading.</p>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">Introduction</h2>
<div class="content has-text-justified">
<p>Autonomous agents leveraging Large Language Models (LLMs) present a transformative approach to decision-making by replicating human processes and workflows across various applications. These systems enhance the problem-solving capabilities of language agents by equipping them with tools and enabling collaboration with other agents, effectively breaking down complex problems into manageable components <cite>Park et al., 2023</cite>, <cite>Havrilla et al., 2024</cite>, <cite>Talebirad et al., 2023</cite>, <cite>Tang et al., 2024</cite>. One prominent application of these autonomous frameworks is in the financial market—a highly complex system influenced by numerous factors, including company fundamentals, market sentiment, technical indicators, and macroeconomic events.</p>
<p>Traditional algorithmic trading systems often rely on quantitative models that struggle to fully capture the complex interplay of diverse factors. In contrast, LLMs excel at processing and understanding natural language data, making them particularly effective for tasks that require textual comprehension, such as analyzing news articles, financial reports, and social media sentiment. Additionally, deep learning-based trading systems often suffer from low explainability, as they rely on hidden features that drive decision-making but are difficult to interpret. Recent advancements in multi-agent LLM frameworks for finance have shown significant promise in addressing these challenges. These frameworks create explainable AI systems, where decisions are supported by evidence and transparent reasoning <cite>Li et al., 2023</cite>, <cite>Wang et al., 2024</cite>, <cite>Yu et al., 2024</cite>, demonstrating their potential in financial applications.</p>
<p>Despite their potential, most current applications of language agents in the financial and trading sectors face two significant limitations:</p>
<strong>Lack of Realistic Organizational Modeling:</strong> Many frameworks fail to capture the complex interactions between agents that mimic the structure of real-world trading firms <cite>Li et al., 2023</cite>, <cite>Wang et al., 2024</cite>, <cite>Yu et al., 2024</cite>. Instead, they focus narrowly on specific task performance, often disconnected from the organizational workflows and established human operating procedures proven effective in trading. This limits their ability to fully replicate and benefit from real-world trading practices.
<strong>Inefficient Communication Interfaces:</strong> Most existing systems use natural language as the primary communication medium, typically relying on message histories or an unstructured pool of information for decision-making <cite>Park et al., 2023</cite>, <cite>Qian et al., 2024</cite>. This approach often results in a "telephone effect", where details are lost, and states become corrupted as conversations lengthen. Agents struggle to maintain context and track extended histories while filtering out irrelevant information from previous decision steps, diminishing their effectiveness in handling complex, dynamic tasks. Additionally, the unstructured pool-of-information approach lacks clear instructions, forcing logical communication and information exchange between agents to depend solely on retrieval, which disrupts the relational integrity of the data.
<p>In this work, we address these key limitations of existing models by introducing a system that overcomes these challenges. First, our framework bridges the gap by simulating the multi-agent decision-making processes typical of professional trading teams. It incorporates specialized agents tailored to distinct aspects of trading, inspired by the organizational structure of real-world trading firms. These agents include fundamental analysts, sentiment/news analysts, technical analysts, and traders with diverse risk profiles. Bullish and bearish debaters evaluate market conditions to provide balanced recommendations, while a risk management team ensures that exposures remain within acceptable limits. Second, to enhance communication, our framework combines structured outputs for control, clarity, and reasoning with natural language dialogue to facilitate effective debate and collaboration among agents. This hybrid approach ensures both precision and flexibility in decision-making.</p>
<p>We validate our framework through experiments on historical financial data, comparing its performance against multiple baselines. Comprehensive evaluation metrics, including cumulative return, Sharpe ratio, and maximum drawdown, are employed to assess its overall effectiveness.</p>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">Related Work</h2>
<h3 class="title is-4">LLMs as Financial Assistants</h3>
<div class="content has-text-justified">
<p>Large Language Models (LLMs) are applied in finance by fine-tuning on financial data or training on financial corpora. This improves the model’s understanding of financial terminology and data, enabling a specialized assistant for analytical support, insights, and information retrieval, rather than trade execution.</p>
<strong>Fine-Tuned LLMs for Finance</strong>
<p>Fine-tuning enhances domain-specific performance. Examples include PIXIU (FinMA) <cite>Xie et al., 2023</cite>, which fine-tuned LLaMA on 136K finance-related instructions; FinGPT <cite>Yang et al., 2023</cite>, which used LoRA to fine-tune models like LLaMA and ChatGLM with about 50K finance-specific samples; and Instruct-FinGPT <cite>Zhang et al., 2023</cite>, fine-tuned on 10K instruction samples from financial sentiment analysis datasets. These models outperform their base versions and other open-source LLMs like BLOOM and OPT <cite>Zhang et al., 2022</cite> in finance classification tasks, even surpassing BloombergGPT <cite>Wu et al., 2023</cite> in several evaluations. However, in generative tasks, they perform similarly or slightly worse than powerful general-purpose models like GPT-4, indicating a need for more high-quality, domain-specific datasets.</p>
<strong>Finance LLMs Trained from Scratch</strong>
<p>Training LLMs from scratch on finance-specific corpora aims for better domain adaptation. Models like BloombergGPT <cite>Wu et al., 2023</cite>, XuanYuan 2.0 <cite>Zhang et al., 2023</cite>, and Fin-T5 <cite>Lu et al., 2023</cite> combine public datasets with finance-specific data during pretraining. BloombergGPT, for instance, was trained on both general and financial text, with proprietary Bloomberg data enhancing its performance on finance benchmarks. These models outperform general-purpose counterparts like BLOOM-176B and T5 in tasks such as market sentiment classification and summarization. While they may not match larger closed-source models like GPT-3 or PaLM <cite>Chowdhery et al., 2022</cite>, they offer competitive performance among similar-sized open-source models without compromising general language understanding.</p>
<p>In summary, finance-specific LLMs developed through fine-tuning or training from scratch show significant improvements in domain-specific tasks, underscoring the importance of domain adaptation and the potential for further enhancements with high-quality finance-specific datasets.</p>
<figure class="image">
<img src="./static/images/schema.png" alt="TradingAgents Overall Framework Organization">
<figcaption class="has-text-centered"><strong>Figure 1:</strong> TradingAgents Overall Framework Organization. <em>I. Analysts Team</em>: Four analysts concurrently gather relevant market information. <em>II. Research Team</em>: The team discusses and evaluates the collected data. <em>III. Trader</em>: Based on the researchers' analysis, the trader makes the trading decision. <em>IV. Risk Management Team</em>: Risk guardians assess the decision against current market conditions to mitigate risks. <em>V. Fund Manager</em>: The fund manager approves and executes the trade.</em></figcaption>
</figure>
</div>
<h3 class="title is-4">LLMs as Traders</h3>
<div class="content has-text-justified">
<p>LLMs act as trader agents making direct trading decisions by analyzing external data like news, financial reports, and stock prices. Proposed architectures include news-driven, reasoning-driven, and reinforcement learning (RL)-driven agents.</p>
<strong>News-Driven Agents</strong>
<p>News-driven architectures integrate stock news and macroeconomic updates into LLM prompts to predict stock price movements. Studies evaluating both closed-source models (e.g., GPT-3.5, GPT-4) and open-source LLMs (e.g., Qwen <cite>Bai et al., 2023</cite>, Baichuan <cite>Yang et al., 2023</cite>) in financial sentiment analysis have shown the effectiveness of simple long-short strategies based on sentiment scores <cite>Lopezlira et al., 2023</cite>. Further research on fine-tuned LLMs like FinGPT and OPT demonstrates improved performance through domain-specific alignment <cite>Unveiling et al.</cite>, <cite>Sentitrade et al.</cite>. Advanced methods involve summarizing news data and reasoning about their relationship with stock prices <cite>Beatunveiling et al.</cite>, <cite>Wang et al., 2024</cite>.</p>
<strong>Reasoning-Driven Agents</strong>
<p>Reasoning-driven agents enhance trading decisions through mechanisms like reflection and debate. Reflection-driven agents, such as FinMem <cite>FinMem et al.</cite> and FinAgent <cite>MultimodalFinMem et al.</cite>, use layered memorization and multimodal data to summarize inputs into memories, inform decisions, and incorporate technical indicators, achieving superior backtest performance while mitigating hallucinations <cite>Ji et al., 2023</cite>. Debate-driven agents, like those in heterogeneous frameworks <cite>Xing et al., 2024</cite> and TradingGPT <cite>Li et al., 2023</cite>, enhance reasoning and factual validity by employing LLM debates among agents with different roles, improving sentiment classification and increasing robustness in trading decisions.</p>
<strong>Reinforcement Learning-Driven Agents</strong>
<p>Reinforcement learning methods align LLM outputs with expected behaviors, using backtesting as rewards. SEP <cite>Koa, 2024</cite> employs RL with memorization and reflection to refine LLM predictions based on market history. Classical RL methods are also used in trading frameworks that integrate LLM-generated embeddings with stock features, trained via algorithms like Proximal Policy Optimization (PPO) <cite>Ding et al., 2023</cite>, <cite>PPO, Year</cite>.</p>
</div>
<h3 class="title is-4">LLMs as Alpha Miners</h3>
<div class="content has-text-justified">
<p>LLMs are also used to generate alpha factors instead of making direct trading decisions. QuantAgent <cite>Wang et al., 2023</cite> demonstrates this by leveraging LLMs to produce alpha factors through an inner-loop and outer-loop architecture. In the inner loop, a writer agent generates a script from a trader's idea, while a judge agent provides feedback. In the outer loop, the code is tested in the real market, and trading results enhance the judge agent. This approach enables progressive approximation of optimal behavior.</p>
<p>Subsequent research, such as AlphaGPT <cite>Wang et al., 2023</cite>, proposes a human-in-the-loop framework for alpha mining with a similar architecture. Both studies showcase the effectiveness of LLM-powered alpha mining systems, highlighting their potential in automating and accelerating the development of trading strategies by generating and refining alpha factors.</p>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">TradingAgents: Role Specialization</h2>
<div class="content has-text-justified">
<p>Assigning LLM agents clear, well-defined roles with specific goals enables the breakdown of complex objectives into smaller, manageable subtasks. Financial trading is a prime example of such complexity, demanding the integration of diverse signals, inputs, and specialized expertise. In the real world, this approach to managing complexity is demonstrated by trading firms that rely on expert teams to collaborate and make high-stakes decisions, underscoring the multifaceted nature of the task.</p>
<p>In a typical trading firm, vast amounts of data are collected, including financial metrics, price movements, trading volumes, historical performance, economic indicators, and news sentiment. This data is then analyzed by quantitative experts (quants), including mathematicians, data scientists, and engineers, using advanced tools and algorithms to identify trends and predict market movements.</p>
<p>Inspired by this organizational structure, <strong>TradingAgents</strong> defines seven distinct agent roles within a simulated trading firm: Fundamentals Analyst, Sentiment Analyst, News Analyst, Technical Analyst, Researcher, Trader, and Risk Manager. Each agent is assigned a specific name, role, goal, and set of constraints, alongside predefined context, skills, and tools tailored to their function. For example, a Sentiment Analyst is equipped with tools like web search engines, Reddit search APIs, X/Twitter search tools, and sentiment score calculation algorithms, while a Technical Analyst can execute code, calculate technical indicators, and analyze trading patterns. More specifically, <strong>TradingAgents</strong> assumes the following teams.</p>
<h3 class="title is-4">Analyst Team</h3>
<div class="content has-text-justified">
<p>The Analyst Team (Figure 2) is composed of specialized agents responsible for gathering and analyzing various types of market data to inform trading decisions. Each agent focuses on a specific aspect of market analysis, bringing together a comprehensive view of the market's conditions.</p>
<figure class="image">
<img src="./static/images/Analyst.png" alt="TradingAgents Analyst Team">
<figcaption class="has-text-centered"><strong>Figure 2:</strong> TradingAgents Analyst Team</figcaption>
</figure>
<ul>
<li><strong>Fundamental Analyst Agents</strong>: These agents evaluate company fundamentals by analyzing financial statements, earnings reports, insider transactions, and other pertinent data. They assess a company's intrinsic value to identify undervalued or overvalued stocks, providing insights into long-term investment potential.</li>
<li><strong>Sentiment Analyst Agents</strong>: These agents process large volumes of social media posts, sentiment scores, and insider sentiments derived from public information and social media activity. They gauge market sentiment to predict how collective investor behavior might impact stock prices in the short term.</li>
<li><strong>News Analyst Agents</strong>: These agents analyze news articles, government announcements, and other macroeconomic indicators to assess the market's macroeconomic state, major world events, and significant company changes. They identify news events that could influence market movements, helping to anticipate sudden shifts in market dynamics.</li>
<li><strong>Technical Analyst Agents</strong>: These agents calculate and select relevant technical indicators, such as Moving Average Convergence Divergence (MACD) and Relative Strength Index (RSI), customized for specific assets. They analyze price patterns and trading volumes to forecast future price movements, assisting in timing entry and exit points.</li>
</ul>
<p>Collectively, the Analyst Team synthesizes data from multiple sources to provide a holistic market analysis. Their combined insights form the foundational input for the Researcher Team, ensuring that all facets of the market are considered in subsequent decision-making processes.</p>
</div>
<h3 class="title is-4">Researcher Team</h3>
<div class="content has-text-justified">
<p>The Researcher Team (Figure 3) is responsible for critically evaluating the information provided by the Analyst Team. Comprised of agents adopting both bullish and bearish perspectives, they engage in multiple rounds of debate to assess the potential risks and benefits of investment decisions.</p>
<figure class="image">
<img src="./static/images/Researcher.png" alt="TradingAgents Researcher Team">
<figcaption class="has-text-centered"><strong>Figure 3:</strong> TradingAgents Researcher Team: Bullish Perspectives and Bearish Perspectives</figcaption>
</figure>
<ul>
<li><strong>Bullish Researchers</strong>: These agents advocate for investment opportunities by highlighting positive indicators, growth potential, and favorable market conditions. They construct arguments supporting the initiation or continuation of positions in certain assets.</li>
<li><strong>Bearish Researchers</strong>: Conversely, these agents focus on potential downsides, risks, and unfavorable market signals. They provide cautionary insights, questioning the viability of investment strategies and highlighting possible negative outcomes.</li>
</ul>
<p>Through this dialectical process, the Researcher Team aims to reach a balanced understanding of the market situation. Their thorough analysis helps in identifying the most promising investment strategies while anticipating possible challenges, thus aiding the Trader Agents in making informed decisions.</p>
</div>
<h3 class="title is-4">Trader Agents</h3>
<div class="content has-text-justified">
<p>Trader Agents (Figure 4) are responsible for executing trading decisions based on the comprehensive analysis provided by the Analyst Team and the nuanced perspectives from the Researcher Team. They assess the synthesized information, considering both quantitative data and qualitative insights, to determine optimal trading actions.</p>
<figure class="image">
<img src="./static/images/Trader.png" alt="TradingAgents Trader Decision-Making Process">
<figcaption class="has-text-centered"><strong>Figure 4:</strong> TradingAgents Trader Decision-Making Process</figcaption>
</figure>
<p>The tasks of <strong>TradingAgents</strong> Trader include:</p>
<ul>
<li>Evaluating recommendations and insights from analysts and researchers.</li>
<li>Deciding on the timing and size of trades to maximize trading returns.</li>
<li>Placing buy or sell orders in the market.</li>
<li>Adjusting portfolio allocations in response to market changes and new information.</li>
</ul>
<p>Trader Agents must balance potential returns against associated risks, making timely decisions in a dynamic market environment. Their actions directly impact the firm's performance, necessitating a high level of precision and strategic thinking.</p>
</div>
<h3 class="title is-4">Risk Management Team</h3>
<div class="content has-text-justified">
<p>The Risk Management Team (Figure 5) monitors and controls the firm's exposure to various market risks. These agents continuously evaluate the portfolio's risk profile, ensuring that trading activities remain within predefined risk parameters and comply with regulatory requirements.</p>
<figure class="image">
<img src="./static/images/RiskMGMT.png" alt="TradingAgents Risk Management Team and Fund Manager Approval Workflow">
<figcaption class="has-text-centered"><strong>Figure 5:</strong> TradingAgents Risk Management Team and Fund Manager Approval Workflow</figcaption>
</figure>
<p>The responsibilities of Risk Management Team include:</p>
<ul>
<li>Assessing factors such as market volatility, liquidity, and counterparty risks.</li>
<li>Implementing risk mitigation strategies, such as setting stop-loss orders or diversifying holdings.</li>
<li>Providing feedback to Trader Agents on risk exposures and suggesting adjustments to trading strategies.</li>
<li>Ensuring that the overall portfolio aligns with the firm's risk tolerance and investment objectives.</li>
</ul>
<p>By offering oversight and guidance, the Risk Management Team helps maintain the firm's financial stability and protect against adverse market events. They play a crucial role in safeguarding assets and ensuring sustainable long-term performance.</p>
<p>All agents in <strong>TradingAgents</strong> follow the ReAct prompting framework <cite>Yao et al., 2023</cite>, which synergizes reasoning and acting. The environment state is shared and monitored by the agents, enabling them to take context-appropriate actions such as conducting research, executing trades, engaging in debates, or managing risks. This design ensures a collaborative, dynamic decision-making process reflective of real-world trading systems.</p>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">TradingAgents: Agent Workflow</h2>
<div class="content has-text-justified">
<h3 class="title is-4">Communication Protocol</h3>
<p>Most existing LLM-based agent frameworks use natural language as the primary communication interface, typically through structured message histories or collections of agent-generated messages <cite>Fatouros et al., 2024</cite>, <cite>Li et al., 2023</cite>, <cite>Yang et al., 2024</cite>, <cite>Yang et al., 2023</cite>. However, relying solely on natural language often proves insufficient for solving complex, long-term tasks that require extensive planning horizons. In such cases, pure natural language communication can resemble a game of telephone—over multiple iterations, initial information may be forgotten or distorted due to context length limitations and an overload of text that obscures critical earlier details <cite>Hong et al., 2024</cite>. To address this limitation, we draw inspiration from frameworks like MetaGPT, which adopt a structured approach to communication. Our model introduces a structured communication protocol to govern agent interactions. By clearly defining each agent's state, we ensure that each role only extracts or queries the necessary information, processes it, and returns a completed report. This streamlined approach reduces unnecessary steps, lowers the risk of message corruption, and keeps interactions focused and efficient, even in complex, long-horizon tasks.</p>
<h3 class="title is-4">Types of Agent Interactions</h3>
<p>In contrast to previous multi-agent trading frameworks, which rely heavily on natural language dialogue, <strong>TradingAgents</strong> agents communicate primarily through structured documents and diagrams. These documents encapsulate the agents' insights in concise, well-organized reports that preserve essential content while avoiding irrelevant information. By utilizing structured reports, agents can query necessary details directly from the global state, eliminating the need for lengthy conversations that risk diluting information, extending the message state indefinitely, and causing data loss. The types of documents and the information they contain are detailed below:</p>
<ul>
<li><strong>Analyst Team</strong>: Fundamental, sentiment, news, and technical analysts compile their research and findings into concise analysis reports specific to their areas of expertise. These reports include key metrics, insights, and recommendations based on their specialized analyses.</li>
<li><strong>Traders</strong>: Traders review and analyze the reports from the analysts, carefully deliberating to produce clear decision signals. They accompany these decisions with detailed reports explaining their rationale and supporting evidence, which are later utilized by the risk management team.</li>
</ul>
<p>Agents engage in natural language dialogue exclusively during agent-to-agent conversations and debates. These concise, focused discussions have been shown to promote deeper reasoning and integrate diverse perspectives, enabling more balanced decisions in complex, long-horizon scenarios—a method particularly relevant to the intricate environment of trading <cite>Du et al., 2023</cite>. This approach seamlessly integrates with our structured framework, as the conversation state is recorded as a structured entry within the overall agent state. The types of communication in these scenarios are detailed below:</p>
<ul>
<li><strong>Researcher Team</strong>: Each researcher agent queries the global agent state for analyst reports and carefully forms their opinion. Two researchers represent opposing perspectives: one bullish and one bearish. They engage in natural language dialogue for $n$ rounds, as determined by the debate facilitator agent. At the conclusion, the facilitator reviews the debate history, selects the prevailing perspective, and records it as a structured entry in the communication protocol.</li>
<li><strong>Risk Management Team</strong>: The risk management team, similar to the researcher team, queries the trader's decision and accompanying report. They then deliberate from three perspectives—risk-seeking, neutral, and risk-conservative—to adjust the trading plan within risk constraints. They engage in $n$ rounds of natural language discussion, guided by a facilitator agent.</li>
<li><strong>Fund Manager</strong>: The fund manager reviews the discussion from the risk management team, determines the appropriate risk adjustments, and updates the trader's decision and report states within the communication protocol.</li>
</ul>
<h3 class="title is-4">Backbone LLMs</h3>
<p>To meet the diverse complexity and speed demands of tasks in our framework, we strategically select Large Language Models (LLMs) based on their strengths. Quick-thinking models, such as <code>gpt-4o-mini</code> and <code>gpt-4o</code>, efficiently handle fast, low-depth tasks like summarization, data retrieval, and converting tabular data to text <cite>OpenAI, 2024</cite>. In contrast, deep-thinking models like <code>o1-preview</code> excel in reasoning-intensive tasks such as decision-making, evidence-based report writing, and data analysis. These models leverage their architectures for multi-round reasoning, producing logically sound, in-depth insights <cite>Zhong et al., 2024</cite>, <cite>Wang et al., 2024</cite>, <cite>OpenAI, 2024</cite>. Additionally, we prioritize models with proven reliability and scalability to ensure optimal performance across various market conditions. We also employ auxiliary expert models for specialized tasks like sentiment analysis.</p>
<p>Specifically, all analyst nodes rely on deep-thinking models to ensure robust analysis, while quick-thinking models handle data retrieval from APIs and tools for efficiency. Researchers and traders use deep-thinking models to generate valuable insights and support well-informed decisions. By aligning the choice of LLMs with the specific requirements of each task, our framework achieves a balance between efficiency and depth of reasoning, which is crucial for effective trading strategies.</p>
<p>This implementation strategy ensures that <strong>TradingAgents</strong> can be deployed without requiring a GPU, relying only on API credits. It also introduces seamless exchangeability of backbone models, enabling researchers to effortlessly replace the model with any locally hosted or API-accessible alternatives in the future. This adaptability supports the integration of improved reasoning models or finance-tuned models customized for specific tasks. As a result, <strong>TradingAgents</strong> is highly scalable and future-proof, offering flexibility to accommodate any backbone model for any of its agents.</p>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">Experiments</h2>
<div class="content has-text-justified">
<p>In this section, we describe the experimental setup used to evaluate our proposed framework. We also provide detailed descriptions of the evaluation metrics employed to assess performance comprehensively.</p>
<figure class="image">
<img src="./static/images/TradingAgents_Performance.png" alt="Performance Comparison">
<figcaption class="has-text-centered"><strong>Table 1:</strong> Performance comparison across all methods using four evaluation metrics. Results highlighted in <strong><span style="color:green;">green</span></strong> represent the best-performing statistic for each model. The improvement row illustrates TradingAgents' performance gains over the top-performing baselines.</figcaption>
</figure>
<h3 class="title is-4">Back Trading</h3>
<p>To simulate a realistic trading environment, we utilize a multi-asset and multi-modal financial dataset comprising of various stocks such as Apple, Nvidia, Microsoft, Meta, Google, and more. The dataset includes:</p>
<ul>
<li><strong>Historical Stock Prices</strong>: Open, high, low, close, volume, and adjusted close prices from January 1st, 2024, to March 29th, 2024.</li>
<li><strong>News Articles</strong>: Daily news updates are gathered from diverse sources such as Bloomberg, Yahoo, EODHD, FinnHub, and Reddit, covering specific company developments, global events, macroeconomic trends, and government updates.</li>
<li><strong>Social Media Posts and Sentiment</strong>: Posts from Reddit, X/Twitter, and other platforms along with sentiment scores of posts calculated by auxiliary language models.</li>
<li><strong>Insider Sentiments and Transactions</strong>: Sentiment derived from public information, including transactions from SEDI and relevant company filings.</li>
<li><strong>Financial Statements and Earnings Reports</strong>: Quarterly and annual reports filed by companies.</li>
<li><strong>Company Profiles and Financial History</strong>: Descriptions of company profiles, target industries, and financial history reported by third parties.</li>
<li><strong>Technical Indicators</strong>: Sixty standard technical analysis indicators calculated for each asset, including MACD, RSI, Bollinger Bands, etc.</li>
</ul>
<h3 class="title is-4">Simulation Setup</h3>
<p>We simulate the trading environment for the period from June 19, 2024, to November 19, 2024. <strong>TradingAgents</strong> facilitates seamless plug-and-play strategies during the simulation, enabling straightforward comparisons with any baseline. Agents make decisions based solely on data available up to each trading day, ensuring no future data is used (eliminating look-ahead bias). Based on their analysis, <strong>TradingAgents</strong> generates trading signals to buy, sell, or hold assets, which are then executed. Afterward, analysis metrics are calculated before proceeding to the next day's data.</p>
<h3 class="title is-4">Baseline Models</h3>
<p>We compare our <strong>TradingAgents</strong> framework against several baselines:</p>
<ul>
<li><strong>Buy and Hold</strong>: Investing equal amounts in all selected stocks and holding them throughout the simulation period.</li>
<li><strong>MACD (Moving Average Convergence Divergence)</strong>: A trend-following momentum strategy that generates buy and sell signals based on the crossover points between the MACD line and signal line.</li>
<li><strong>KDJ and RSI (Relative Strength Index)</strong>: A momentum strategy combining KDJ (stochastic oscillator) and RSI (relative strength index) indicators to identify overbought and oversold conditions for trading signals.</li>
<li><strong>ZMR (Zero Mean Reversion)</strong>: A mean reversion trading strategy that generates signals based on price deviations from and subsequent reversions to a zero reference line.</li>
<li><strong>SMA (Simple Moving Average)</strong>: A trend-following strategy that generates trading signals based on crossovers between short-term and long-term moving averages.</li>
</ul>
<h3 class="title is-4">Evaluation Metrics</h3>
<figure class="image">
<img src="./static/images/TradingAgents_CR_AAPL.png" alt="Cumulative Returns on AAPL">
<figcaption class="has-text-centered"><strong>Figure 6:</strong> TradingAgents: Cumulative Returns (CR) and Detailed Transaction History for AAPL.</figcaption>
</figure>
<p>To thoroughly evaluate the performance of our <strong>TradingAgents</strong> framework, we use widely recognized metrics to assess the risk management, profitability, and safety of the TradingAgents strategy in comparison to baseline approaches. Here we describe these metrics:</p>
<h4 class="title is-5">Cumulative Return (CR)</h4>
<p>The cumulative return measures the total return generated over the simulation period. It is calculated as:</p>
<p>
<strong>CR</strong> = ((V<sub>end</sub> - V<sub>start</sub>) / V<sub>start</sub>) × 100%
</p>
<p>where V<sub>end</sub> is the portfolio value at the end of the simulation, and V<sub>start</sub> is the initial portfolio value.</p>
<h4 class="title is-5">Annualized Return (AR)</h4>
<p>The annualized return normalizes the cumulative return over the number of years:</p>
<p>
<strong>AR</strong> = (((V<sub>end</sub> / V<sub>start</sub>)^(1/N)) - 1) × 100%
</p>
<p>where N is the number of years in the simulation.</p>
<h4 class="title is-5">Sharpe Ratio (SR)</h4>
<p>The Sharpe ratio measures risk-adjusted return by comparing a portfolio's excess return over the risk-free rate to its volatility:</p>
<p>
<strong>SR</strong> = (R̄ - R<sub>f</sub>) / σ
</p>
<p>where R̄ is the average portfolio return, R<sub>f</sub> is the risk-free rate (e.g., yield of 3-month Treasury bills), and σ is the standard deviation of the portfolio returns.</p>
<h4 class="title is-5">Maximum Drawdown (MDD)</h4>
<p>Maximum drawdown measures the largest peak-to-trough decline in the portfolio value:</p>
<p>
<strong>MDD</strong> = max<sub>t ∈ [0, T]</sub> ((Peak<sub>t</sub> - Trough<sub>t</sub>) / Peak<sub>t</sub>) × 100%
</p>
<figure class="image">
<img src="./static/images/TradingAgents_ROUGE_AAPL.png" alt="ROUGE Score Comparison">
<figcaption class="has-text-centered"><strong>Figure 7:</strong> ROUGE Score Comparison</figcaption>
</figure>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">Results and Analysis</h2>
<div class="content has-text-justified">
<h3 class="title is-4">Performance Comparison</h3>
<h4 class="title is-5">Cumulative and Annual Returns</h4>
<p>Table 1 and Figures 6, 7, and 8 highlight that our method significantly outperforms existing rule-based trading baselines, particularly in profitability, as measured by returns. <strong>TradingAgents</strong> achieves at least a 23.21% cumulative return and 24.90% annual return on the three sampled stocks, outperforming the best-performing baselines by a margin of at least 6.1%. Notably, on the AAPL stock—a particularly challenging case due to market volatility during the testing period—traditional methods struggled, as their patterns failed to generalize to this situation. In contrast, <strong>TradingAgents</strong> excelled even under these adverse conditions, achieving returns exceeding 26% within less than three months.</p>
<h4 class="title is-5">Sharpe Ratio</h4>
<p>The Sharpe Ratio performance highlights <strong>TradingAgents</strong>' exceptional ability to deliver superior risk-adjusted returns, consistently outperforming all baseline models across AAPL, GOOGL, and AMZN with Sharpe Ratios of at least 5.60—surpassing the next best models by a significant margin of at least 2.07 points. This result underscores <strong>TradingAgents</strong>' effectiveness in balancing returns against risk, a critical metric for sustainable and predictable investment growth. By excelling over market benchmarks like Buy-and-Hold and advanced strategies such as KDJRSI, SMA, MACD, and ZMR, <strong>TradingAgents</strong> demonstrates its adaptability and robustness in diverse market conditions. Its ability to maximize returns while maintaining controlled risk exposure establishes a solid foundation for multi-agent and debate-based automated trading algorithms.</p>
<h4 class="title is-5">Maximum Drawdown</h4>
<p>While rule-based baselines demonstrated superior performance in controlling risk, as reflected by their maximum drawdown scores, they fell short in capturing high returns. This trade-off between risk and reward underscores <strong>TradingAgents</strong>' strength as a balanced approach. Despite higher returns being typically associated with higher risks, <strong>TradingAgents</strong> maintained a relatively low maximum drawdown compared to many baselines. Its effective risk-control mechanisms, facilitated by the debates among risk-control agents, ensured that the maximum drawdown remained within a manageable limit, not exceeding 2%. This demonstrates <strong>TradingAgents</strong>' capability to strike a robust balance between maximizing returns and managing risk effectively.</p>
<h4 class="title is-5">Explainability</h4>
<p>A significant drawback of current deep learning methods for trading is their dense and complex architectures, which often render the decisions made by trading agents indecipherable to humans. This challenge, rooted in the broader issue of AI explainability, is particularly critical for trading agents, as they operate in real-world financial markets, often involving substantial sums of money where incorrect decisions can lead to severe consequences and losses.</p>
<p>In contrast, an LLM-based agentic framework for trading offers a transformative advantage: its operations and decisions are communicated in natural language, making them highly interpretable to humans. To illustrate this, we provide the full trading log of <strong>TradingAgents</strong> for a single day in the Appendix, showcasing its use of the ReAct-style prompting framework <cite>Yao et al., 2023</cite>. Each decision made by the agents is accompanied by detailed reasoning, tool usage, and thought processes, enabling traders to easily understand and debug the system. This transparency empowers traders to fine-tune and adjust the framework to account for factors influencing decisions, offering a significant edge in explainability over traditional deep-learning trading algorithms.</p>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full-width">
<h2 class="title is-3">Conclusion</h2>
<div class="content has-text-justified">
<p>In this paper, we introduced <strong>TradingAgents</strong>, an LLM-agent-powered stock trading framework that simulates a realistic trading firm environment with multiple specialized agents engaging in agentic debates and conversations. Leveraging the capabilities of LLMs to process and analyze diverse data sources, the framework enables informed trading decisions while utilizing multi-agent interactions to enhance performance through comprehensive reasoning and debate before acting. By integrating agents with distinct roles and risk profiles, along with a reflective agent and a dedicated risk management team, <strong>TradingAgents</strong> significantly improves trading outcomes and risk management compared to baseline models. Additionally, the collaborative nature of these agents ensures adaptability to varying market conditions. Extensive experiments demonstrate that <strong>TradingAgents</strong> outperforms traditional trading strategies and baselines in cumulative return, Sharpe ratio, and other critical metrics. Future work will focus on deploying the framework in a live trading environment, expanding agent roles, and incorporating real-time data processing to enhance performance further.</p>
</div>
</div>
</div>
</div>
</section>
<footer class="footer">
<div class="container">
<div class="content has-text-centered">
<a class="icon-link" href="https://arxiv.org/abs/XXXX.XXXXX"><i class="fas fa-file-pdf"></i></a>
<a class="icon-link" href="https://github.com/Yijia-Xiao/TradingAgents"><i class="fab fa-github"></i></a>
</div>
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>This website is licensed under a <a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/">Creative Commons Attribution-ShareAlike 4.0 International License</a>.</p>
</div>
</div>
</div>
</div>
</footer>
</body>
</html>