Paper: arXiv 2403.10482

Authors: Bruno de Melo, Jamiel Sheikh

Abstract

Performance attribution analysis, defined as the process of explaining the drivers of the excess performance of an investment portfolio against a benchmark, stands as a significant feature of portfolio management and plays a crucial role in the investment decision-making process, particularly within the fund management industry. Rooted in a solid financial and mathematical framework, the importance and methodologies of this analytical technique are extensively documented across numerous academic research papers and books. The integration of large language models (LLMs) and AI agents marks a groundbreaking development in this field. These agents are designed to automate and enhance the performance attribution analysis by accurately calculating and analyzing portfolio performances against benchmarks. In this study, we introduce the application of an AI Agent for a variety of essential performance attribution tasks, including the analysis of performance drivers and utilizing LLMs as calculation engine for multi-level attribution analysis and question-answering (QA) tasks. Leveraging advanced prompt engineering techniques such as Chain-of-Thought (CoT) and Plan and Solve (PS), and employing a standard agent framework from LangChain, the research achieves promising results: it achieves accuracy rates exceeding 93% in analyzing performance drivers, attains 100% in multi-level attribution calculations, and surpasses 84% accuracy in QA exercises that simulate official examination standards. These findings affirm the impactful role of AI agents, prompt engineering and evaluation in advancing portfolio management processes, highlighting a significant development in the practical application and evaluation of Generative AI technologies within the domain.

Complexity vs Empirical Score

  • Math Complexity: 2.5/10
  • Empirical Rigor: 4.0/10
  • Quadrant: Philosophers — conceptual discussion, lower math and empirical scores

Why this score: The paper’s mathematics is straightforward, relying on established formulas like the Brinson-Fachler model without dense derivations, while empirical rigor is limited to accuracy metrics on simulated tasks using an AI agent, lacking backtesting or real-market data implementation.

Research Flowchart

  flowchart TD
  A["Research Goal:<br>Can GPT-4 powered AI Agent<br>perform Performance Attribution Analysis?"] --> B

  subgraph B ["Methodology & Inputs"]
      B1["Prompt Engineering<br>CoT & Plan and Solve"] --> C
      B2["Data: Portfolio vs Benchmark<br>Multi-level Data"] --> C
      B3["Agent Framework:<br>LangChain"] --> C
  end

  C["Computational Processes:<br>Performance Driver Analysis<br>Multi-level Attribution Calculation<br>QA Simulation"] --> D

  subgraph D ["Key Findings / Outcomes"]
      D1[">93% Accuracy<br>Performance Drivers"]
      D2["100% Accuracy<br>Multi-level Calculation"]
      D3[">84% Accuracy<br>QA (Exam Standard)"]
  end

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 using an AI agent powered by GPT‑4 to automate performance attribution tasks, including driver analysis, multi‑level calculations and QA.

Source sentence
In this study, we introduce the application of an AI Agent for a variety of essential performance attribution tasks, including the analysis of performance drivers and utilizing LLMs as calculation engine for multi-level attribution analysis and question-answering (QA) tasks.

Assumes 2

  • The AI Agent is used “as is” with no code modifications. model
    Source sentence
    One important premise underlying the hypothesis and the research questions is that the AI Agent is used “as is”, meaning that there is no change to its original code, despite being an open-source project.
  • The dataset follows a standard long‑only equity portfolio attribution format. data
    Source sentence
    In practical terms, a standard performance attribution dataset for long-only equity portfolio might contains columns like: Breakdown by GICS Sector or Currency, Portfolio and Benchmark Weights, Portfolio and Benchmark Returns, Variation in…

Data

  • standard performance attribution dataset for long‑only equity portfolio benchmark unknown
    Source sentence
    In practical terms, a standard performance attribution dataset for long-only equity portfolio might contains columns like Breakdown by GICS Sector or Currency, Portfolio and Benchmark Weights, Portfolio and Benchmark Returns, Variation in…

Costs

Not applicable The paper reports no trading or execution result.

Finds 3

  • The AI agent achieved accuracy rates exceeding 93% in analyzing performance drivers. performance
    Source sentence
    it achieves accuracy rates exceeding 93% in analyzing performance drivers
  • The agent attained 100% accuracy in multi‑level attribution calculations. performance
    Source sentence
    attains 100% in multi-level attribution calculations
  • The agent surpassed 84% accuracy in QA exercises simulating official examination standards. performance
    Source sentence
    surpasses 84% accuracy in QA exercises that simulate official examination standards

Checked by others

No other paper in the archive has been found to support or contradict these findings (compared with 14 related papers). Coverage is partial, so this is neither replication nor refutation.

Cited by 2 papers in the archive (2 by independent authors)

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 (6)

  1. The Rise and Potential of Large Language Model Based Agents: A Survey (2023) — link
  2. Gorilla: Large Language Model Connected with Massive APIs (2023) — link
  3. Self-Consistency Improves Chain of Thought Reasoning in Language Models (2022) — link
  4. Allocation Effect being positive and Selection Effect being negative, implies that Allocation is less than the absolute value of Selection, OR
  5. viii. Second term is positive: Benchmark sector return is greater than the overall Benchmark return, or benchmark sector outperformed the overall market
  6. ii. Portfolio sector return is less than the Benchmark sector return, or portfolio sector underperformed the benchmark sector

0 of 6 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.