Identify information requirements in real-world hypergraph benchmarks

Determine which information—higher-order, lower-order, feature-based, or combinations thereof—is needed to achieve a given level of predictive performance on real-world hypergraph-learning benchmarks.

Background

The paper distinguishes between demonstrating that a higher-order model performs well and establishing that higher-order information is necessary for that performance. Synthetic controls can identify information requirements by construction, but real-world benchmarks contain heterogeneous features, relational structures, and prediction targets whose dependencies are not known beforehand.

The unresolved problem is to characterize the information requirements of real-world hypergraph-learning benchmarks. The authors suggest that semi-synthetic benchmarks could help address this gap by preserving realistic features and relational structure while introducing prediction targets with controlled dependence on higher-order information.

References

Real-world benchmarks offer breadth and realistic heterogeneity, but we generally do not know in advance which information is needed to achieve a given level of predictive performance.

— Do Higher-Order Models Win for Higher-Order Reasons? Rethinking Performance Gains in Hypergraph Learning  (2610.07981 - Bu et al., 6 Oct 2026) in Section 6, “Limitations and Future Directions,” paragraph “Benchmarks with known information requirements”