Determine whether COOBA’s failure mode generalizes beyond Python

Determine whether COOBA’s AST-based adversarial cross-project bug-localization failure mode replicates in Java and other programming languages, thereby establishing whether the observed behavior generalizes beyond Python projects.

Background

The empirical evaluation covers 13 open-source Python projects, and COOBA’s AST-based encoder may depend on language-specific properties such as identifier namespaces, call-graph depth, and type hierarchies. Because these properties differ across languages, the observed cold-start and transfer behavior may not extend to Java or other programming-language ecosystems. The paper explicitly identifies this generalization question as unresolved.

References

The mechanism-based explanations (Section~\ref{sec:architecture}) are grounded in inference-time properties that are language-agnostic in principle, but COOBA's AST-based encoder is sensitive to identifier namespaces, call-graph depth, and type hierarchies that differ between Python and Java, so whether its failure mode replicates there is open.

— Cross-Project Bug Localization: What Works, What Doesn't, and Why  (2610.17578 - Rao et al., 13 Oct 2026) in Threats to Validity, External validity, Section 5.4