Reproduce “FinEvolveKG: Controlled and Attributable Prompt Evolution for Financial Knowledge Graph Construction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Abhinav Arun, Jayita Bhattacharyya, Fabrizio Dimino, Bhaskarjit Sarmah, Stefano Pasquali
FinEvolveKG separates shared financial knowledge-graph rules from persona-specific extraction instructions, and optimizes the two blocks in sequence while keeping model weights fixed. On ticker-disjoint held-out 2024 IT filings, its blended scores exceed the frozen baseline and tested full-prompt workflow across three analyst personas. These workflow comparisons do not establish that block separation caused the gains or guarantees unchanged behavior.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.