Paper: arXiv 2405.14767

Authors: Hongyang Yang, Boyu Zhang, Neng Wang, Cheng Guo, Xiaoli Zhang, Likun Lin, Junlin Wang, Tianyu Zhou, Mao Guan, Runjia Zhang, Christina Dan Wang

Abstract

As financial institutions and professionals increasingly incorporate Large Language Models (LLMs) into their workflows, substantial barriers, including proprietary data and specialized knowledge, persist between the finance sector and the AI community. These challenges impede the AI community’s ability to enhance financial tasks effectively. Acknowledging financial analysis’s critical role, we aim to devise financial-specialized LLM-based toolchains and democratize access to them through open-source initiatives, promoting wider AI adoption in financial decision-making. In this paper, we introduce FinRobot, a novel open-source AI agent platform supporting multiple financially specialized AI agents, each powered by LLM. Specifically, the platform consists of four major layers: 1) the Financial AI Agents layer that formulates Financial Chain-of-Thought (CoT) by breaking sophisticated financial problems down into logical sequences; 2) the Financial LLM Algorithms layer dynamically configures appropriate model application strategies for specific tasks; 3) the LLMOps and DataOps layer produces accurate models by applying training/fine-tuning techniques and using task-relevant data; 4) the Multi-source LLM Foundation Models layer that integrates various LLMs and enables the above layers to access them directly. Finally, FinRobot provides hands-on for both professional-grade analysts and laypersons to utilize powerful AI techniques for advanced financial analysis. We open-source FinRobot at \url{“https://github.com/AI4Finance-Foundation/FinRobot"}.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 6.0/10
  • Quadrant: Street Traders — empirical and applied, lower math complexity

Why this score: The paper introduces a modular platform architecture with operational layers rather than heavy mathematical derivations, resulting in moderate complexity. It demonstrates practical implementation with a public GitHub repository and specific evaluation cases, giving it solid empirical grounding for application development.

Research Flowchart

  flowchart TD
  A["Research Goal: Develop Financial-Specialized LLM Platform"] --> B["Key Methodology: Four-Layer Architecture"]
  
  subgraph B ["Four-Layer Architecture"]
      B1["Financial AI Agents<br>Financial CoT Formulation"]
      B2["Financial LLM Algorithms<br>Dynamic Model Configuration"]
      B3["LLMOps & DataOps<br>Model Training/Fine-tuning"]
      B4["Multi-source LLM Foundation<br>Integrated LLM Access"]
  end
  
  B --> C["Computational Process: Pipeline Execution"]
  C --> D["Key Outcomes: FinRobot Platform"]
  
  D --> D1["Open-Source AI Agent Platform<br>github.com/AI4Finance-Foundation/FinRobot"]
  D --> D2["Specialized Financial Analysis<br>For Professional & Layperson Users"]
  D --> D3["Democratized Access to Financial AI Tools"]

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 FinRobot, an open-source AI agent platform with four layers that supports multiple financially specialized AI agents powered by large language models.

Source sentence
In this paper, we introduce FinRobot, a novel open-source AI agent platform supporting multiple financially specialized AI agents, each powered by LLM.

Assumes

No explicit assumptions found in the text.

Data

  • Finnhub · Finnhub text-news public
    Source sentence
    This section will give a solid overview of the Market Forecaster targeted at the US stock market and Chinese stock market with data sources Finnhub and EastMoney.
  • EastMoney · EastMoney text-news public
    Source sentence
    This section will give a solid overview of the Market Forecaster targeted at the US stock market and Chinese stock market with data sources Finnhub and EastMoney.
  • DOW 30 market-prices public
    U.S. large-cap stocks
    Source sentence
    Model This AI Agent is powered by fine-tuned FinGPT-Forecasters using Llama-2-7b-chat-hf with LoRA, leveraging data from the latest year’s DOW 30 for the US Market and SSE 50 for the Chinese Market to ensure precise forecasts for these ma-jor stocks and demonstrating robust generalization abilities across various stock symbols.
  • SSE 50 market-prices public
    Chinese large-cap stocks
    Source sentence
    Model This AI Agent is powered by fine-tuned FinGPT-Forecasters using Llama-2-7b-chat-hf with LoRA, leveraging data from the latest year’s DOW 30 for the US Market and SSE 50 for the Chinese Market to ensure precise forecasts for these ma-jor stocks and demonstrating robust generalization abilities across various stock symbols.

Costs

Not applicable The paper reports no trading or execution result.

Finds

No quantified findings extracted.

Checked by others

No other paper in the archive has been found to support or contradict these findings. Coverage is partial, so this is neither replication nor refutation.

Cited by 19 papers in the archive (15 by independent authors)

Code

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

Cited References (85)

  1. Language models are few-shot learners (2020)
  2. BloombergGPT : A large language model for finance (2023) — link
  3. FinRL-Meta : Market Environments and Benchmarks for Data-Driven Financial Reinforcement Learning (2022)
  4. Dynamic Datasets and Market Environments for Financial Reinforcement Learning (2024)
  5. FinRL : Deep reinforcement learning framework to automate trading in quantitative finance (2021)
  6. FinRL : A deep reinforcement learning library for automated stock trading in quantitative finance (2020)
  7. A Practical Machine Learning Approach for Dynamic Stock Recommendation (2018) — link
  8. LoRA : Low-rank adaptation of large language models (2021)
and 77 more
  1. QLoRA : Efficient Finetuning of Quantized LLMs (2023) — link
  2. Finbert: Financial sentiment analysis with pre-trained language models (2019) — link
  3. Plato-xl: Exploring the large-scale pre-training of dialogue generation (2021) — link
  4. Bernice: a multilingual pre-trained encoder for Twitter (2022)
  5. How twitter is changing the nature of financial news discovery (2016)
  6. Pretrained language models for biomedical and clinical tasks: understanding and extending the state-of-the-art (2020)
  7. Bert: Pre-training of deep bidirectional transformers for language understanding (2018) — link
  8. GPT-3: Its nature, scope, limits, and consequences (2020)
  9. How contextual are contextualized word representations? comparing the geometry of BERT, ELMo, and GPT-2 embeddings (2019) — link
  10. Attention is All you Need (2017) — link
  11. Improving language understanding by generative pre-training (2018)
  12. Bart: Denoising sequence-to-sequence pre-training for natural language generation, translation, and comprehension (2019) — link
  13. Lamda: Language models for dialog applications (2022) — link
  14. Good debt or bad debt: Detecting semantic orientations in economic texts (2014)
  15. Www'18 open challenge: financial opinion mining and question answering (2018)
  16. Training language models to follow instructions with human feedback (2022)
  17. When flue meets flang: Benchmarks and large pre-trained language model for financial domain (2022) — link
  18. FinRED: A Dataset for Relation Extraction in Financial Domain (2022) — link
  19. Finqa: A dataset of numerical reasoning over financial data (2021) — link
  20. FinTral: A Family of GPT-4 Level Multimodal Financial Large Language Models (2024) — link
  21. DocLLM: A layout-aware generative language model for multimodal document understanding (2023) — link
  22. F in R ead: A Transfer Learning Based Tool to Assess Readability of Definitions of Financial Terms (2021) — link
  23. Enhancing Financial Sentiment Analysis via Retrieval Augmented Large Language Models — → 2310.04027
  24. Instruct-FinGPT: Financial Sentiment Analysis by Instruction Tuning of General-Purpose Large Language Models — → 2306.12659
  25. The rise and potential of large language model based agents: A survey (2023) — link
  26. Large language models empowered agent-based modeling and simulation: A survey and perspectives (2023) — link
  27. A survey on large language model based autonomous agents (2023) — link
  28. FinAgent: A Multimodal Foundation Agent for Financial Trading: Tool-Augmented, Diversified, and Generalist — → 2402.18485
  29. FinMem: A Performance-Enhanced LLM Trading Agent with Layered Memory and Character Design — → 2311.13743
  30. Musicagent: An ai agent for music understanding and generation with large language models (2023) — link
  31. Large language models in medicine (2023)
  32. Medagents: Large language models as collaborators for zero-shot medical reasoning (2023) — link
  33. Taskweaver: A code-first agent framework (2023) — link
  34. Codepori: Large scale model for autonomous software development by using multi-agents (2024) — link
  35. Can a GPT4-Powered AI Agent Be a Good Enough Performance Attribution Analyst? — → 2403.10482
  36. Voyager: An open-ended embodied agent with large language models (2023) — link
  37. AutoGPT (2023)
  38. LangChain (2022) — link
  39. AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation Framework (2023) — link
  40. MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework (2023) — link
  41. Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face (2024)
  42. Cognitive architectures for language agents (2023) — link
  43. Generative agents: Interactive simulacra of human behavior (2023)
  44. Communicative Agents for Software Development (2023) — link
  45. Dify (2024)
  46. Large language models for robotics: A survey (2023) — link
  47. Agent ai: Surveying the horizons of multimodal interaction (2024) — link
  48. Financial intelligence, revised edition: A manager’s guide to knowing what the numbers really mean (2013)
  49. Sentiment analysis algorithms and applications: A survey (2014)
  50. Literature review: Machine learning techniques applied to financial market prediction (2019)
  51. Deep learning for stock market prediction (2020)
  52. A systematic review of stock market prediction using machine learning and statistical techniques (2022)
  53. Applications of deep learning in stock market prediction: recent progress (2021)
  54. FinGPT: Open-Source Financial Large Language Models — → 2306.06031
  55. Big Data: Deep Learning for financial sentiment analysis (2018)
  56. FinBERT: A large language model for extracting information from financial text (2023)
  57. Value Investing: From Graham to Buffett and Beyond (2004) — link
  58. Accounting for Value (2010) — link
  59. Chain-of-thought prompting elicits reasoning in large language models (2022)
  60. Ai in finance: challenges, techniques, and opportunities (2022)
  61. Artificial intelligence techniques in finance and financial markets: a survey of the literature (2021)
  62. Temporal Data Meets LLM–Explainable Financial Time Series Forecasting — → 2306.11025
  63. FinGPT: Instruction Tuning Benchmark for Open-Source Large Language Models in Financial Datasets — → 2310.04793
  64. Data-centric FinGPT: Democratizing Internet-scale Data for Financial Large Language Models (2023)
  65. Technical analysis of the financial markets: A comprehensive guide to trading methods and applications (1999)
  66. Fundamental analysis, future earnings, and stock prices (1997)
  67. Application of big data with fintech in financial services (2021)
  68. Harnessing LLMs for temporal data-a study on explainable financial time series forecasting (2023)
  69. Deep reinforcement learning for automated stock trading: An ensemble strategy (2020)
  70. LLaMA 2 : Open Foundation and Fine-Tuned Chat Models (2023)
  71. Glm-130b: An open bilingual pre-trained model (2022) — link
  72. Falcon-40B : an open large language model with state-of-the-art performance (2023)
  73. Beyond Chain-of-Thought: A Survey of Chain-of-X Paradigms for LLMs (2024) — link
  74. What Are Tools Anyway? A Survey from the Language Model Perspective (2024) — link
  75. MTEB: Massive Text Embedding Benchmark (2022) — link
  76. ResearchAgent: Iterative Research Idea Generation over Scientific Literature with Large Language Models (2024) — link
  77. Financial Statement Analysis with Large Language Models — → 2407.17866

9 of 85 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.