{"paper_id":"2407.10909","title":"FinDKG: Dynamic Knowledge Graphs with Large Language Models for Detecting Global Trends in Financial Markets","authors":["Xiaohui Victor Li","Francesco Sanna Passino"],"paper_date":"2024-07-15","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2407.10909v2","page_url":"https://thequant.space/flowcharts/findkg-dynamic-knowledge-graphs-with-large-language-models/","scores":{"empirical_rigor":6,"math_complexity":7.5,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes an open‑source fine‑tuned LLM (ICKG) to generate a financial dynamic knowledge graph (FinDKG) and an attention‑based GNN (KGTransformer) for link prediction and thematic investing.","quote":"an attention‑based GNN architecture for analysing it, called KGTransformer.","summary":"Dynamic knowledge graphs (DKGs) are popular structures to express different types of connections between objects over time. They can also serve as an efficient mathematical tool to represent information extracted from complex unstructured data sources, such as text or images. Within financial…","strategy":"FinDKG-AI (KGTransformer-predicted AI Portfolio)"},"assumptions":{"status":"available","items":[{"assumption":"A mapping function τ from each entity to an entity type exists.","type":"model","stated":"explicit","quote":"we assume that a function τ: E → C_E exists, mapping each entity to an entity type"},{"assumption":"Model training uses three distinct random seeds and results are averaged over runs.","type":"model","stated":"explicit","quote":"we employ three distinct random seeds, shared across different models. The final results are reported as averages over these training runs"},{"assumption":"The portfolio is monthly‑rebalanced, long‑only, and weights sum to 100%.","type":"execution","stated":"explicit","quote":"This selection forms the basis of a monthly‑rebalanced, AI‑focused long‑only portfolio within the US S&P 500. The portfolio is constructed by using the normalised predicted likelihood scores as the holding weight, which sum up to 100%"},{"assumption":"All experiments are run on a single NVIDIA A100 GPU with 40 GB memory.","type":"execution","stated":"explicit","quote":"Both model training and evaluations are consistently conducted on an identical computational environment: a single NVIDIA A100 GPU cloud server with 40GB of memory"}]},"datasets":{"status":"available","items":[{"name":"YAGO","provider":null,"kind":"benchmark","access":"public","period":null,"universe":null,"quote":"YAGO 161,540 19,523 20,026 10,623 10 -"},{"name":"WIKI","provider":null,"kind":"benchmark","access":"public","period":null,"universe":null,"quote":"WIKI 539,286 67,538 63,110 12,554 24 -"},{"name":"ICEWS14","provider":null,"kind":"benchmark","access":"public","period":null,"universe":null,"quote":"ICEWS14 275,367 48,528 341,409 12,498 260 -"},{"name":"FinDKG","provider":null,"kind":"text-news","access":"public","period":null,"universe":null,"quote":"FinDKG 119,549 11,444 13,069 13,645 15 12"}],"known":[],"sample":{"start":"2022-06","end":"2023-12","market":"S&P 500 stocks","asset_class":"equities","frequency":"monthly"},"data_resolution":"weekly-monthly"},"transaction_costs":{"status":"available","modeled":"no","headline_net_of_costs":false,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"no","evidence_card":"no","evidence_card_quoted":false},"meaning":"a trading or execution result is reported without costs"},"results":{"status":"available","evaluation_type":"trading-pnl","findings":[{"finding":"KGTransformer achieves the highest MRR and Hits@3/10 on YAGO and WIKI, outperforming prior models.","kind":"predictive-accuracy","quote":"KGTransformer outperforms competitors on the YAGO and WIKI datasets"},{"finding":"On FinDKG, incorporating entity types yields roughly a 10% improvement in MRR and Hits@3/10 over temporal baselines.","kind":"predictive-accuracy","quote":"Integrating these types into the KGTransformer enhances performance significantly, resulting in an approximate 10% improvement in MRR and Hits@3,10 metrics over temporal baselines"},{"finding":"The FinDKG‑AI portfolio (KGTransformer‑based) attains the highest annualized return (39.6%) and Sharpe ratio (1.81) among all tested portfolios.","kind":"performance","quote":"FinDKG‑AI achieves the highest annualized return and Sharpe ratio across all portfolios"},{"finding":"FinDKG‑AI outperforms the EvoKG‑based AI portfolio, reflecting superior link‑prediction capability of KGTransformer.","kind":"performance","quote":"The FinDKG‑AI portfolio also outperforms the EvoKG‑based strategy, aligning with the better link prediction capabilities of the KGTransformer model architecture"}],"reported_performance":{"strategy":"FinDKG-AI (KGTransformer-predicted AI Portfolio)","sharpe":1.81,"sharpe_benchmark":null,"annual_return_pct":39.6,"total_return_pct":null,"max_drawdown_pct":18.2},"out_of_sample":{"method":"walk-forward","headline_out_of_sample":true}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":32,"citations":{"cited_by_in_corpus":3,"by_independent_authors":3,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":2,"unclassified":1},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":true,"code_url":"https://huggingface.co/victorlxh/ICKG-v3","note":null},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T18:06:13Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}