Reproduce “TIEM: Temporal Integration of Hypergraph Evidence and Skill Memory for Event-Driven Financial Forecasting” 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
Wenjin Liu, Shen Pang, Chenxi Wang, Tiesunlong Shen, Jiajie He, Zhe Cui, Xiaobao Wu, Anh Tuan Luu and 1 more
TIEM is a temporally constrained framework for forecasting post-event stock movements by combining timestamped hypergraph retrieval, provenance-aware case memory, and joint LLM reasoning. Across five benchmarks and two language models, it outperformed the evaluated baselines on accuracy, MCC, and macro F1 while passing a 1,280-case temporal audit with no detected look-ahead violations.
Key insights
- Temporal controls are not merely an auditing feature: removing them caused larger accuracy drops than removing the hypergraph, memory, or fusion components in the reported Astock ablations.
- Multi-scale organization helps connect a focal catalyst to same-day developments, earlier company-specific events, and cross-stock or industry-wide themes without discarding provenance.
- The Name-Date Probe found no reliable above-base-rate performance from stock identifiers and dates alone on Astock and FinPURE, while adding focal-event text provided a positive content contribution on Astock.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.