Measure residue dependence on schema field overlap

Determine how the H5 borrowed-signature residue varies as a function of field-name and type overlap among tools in the registry.

Background

The controlled experiments suggest that schema-indistinguishable H5 errors may become more common when multiple tools share field names and types, but the paper explicitly describes this as a correlational observation rather than an established mechanism.

A controlled sweep over the degree of field overlap is needed to determine whether, and how, schema similarity causes the H5 residue rate to change. This would clarify the external validity of the synthetic benchmark results.

References

Absolute hallucination rates and the exact H5 residue will differ on real tool catalogs, particularly the residue, which we observe to co-vary with how many exposed tools share field names and types (a correlational observation, not a demonstrated mechanism; a controlled field-overlap sweep is left to future work), and which is a sample estimate (the controlled ablation and the released HTB suite draw independent samples of the same borrowed-signature construction and report distinct but comparable residues, $0.1525$ in the controlled ablation and $0.1925$ on the HTB single-registry split, both entirely schema-valid).

— Closed-World Resolution Against Tool Hallucination in LLM Agents  (2609.19425 - Iyer, 16 Sep 2026) in Section 5, subsection “External validity”; also Section Limitations and Future Work