LR-Robot: An Human-in-the-Loop LLM Framework for Systematic Literature Reviews with Applications in Financial Research

Battle Count: 50

The paper is primarily about methodology for systematic literature reviews and academic classification, not about trading strategies or quantitative models directly. However, it provides valuable insights into the option pricing literature landscape, identifies emerging trends (ML approaches growing to ~25%), and could help researchers quickly identify relevant papers for developing trading strategies. The framework itself is a meta-tool for research rather than a trading application.

The framework requires: (1) domain expertise to design taxonomies and prompt constraints, (2) access to multiple LLM APIs for evaluation, (3) iterative human-in-the-loop evaluation cycles, (4) RAG knowledge base construction, (5) citation network analysis with PageRank. The core classification pipeline is relatively straightforward (prompt + LLM call), but the full framework with evaluation, refinement, and downstream analyses requires significant engineering effort. No code is publicly released.

Data and code available upon request. The framework uses publicly available Scopus database queries and standard LLM APIs. Prompt designs are fully documented in Appendix B. However, the iterative human-in-the-loop evaluation process and expert taxonomy design are inherently difficult to reproduce exactly. The specific Scopus query is provided, and evaluation metrics are clearly defined.