Determine generalization to unseen dialogue domains

Determine whether BDI-based mentalized adversarial self-play generalizes beyond the evaluated persuasion, emotional-support, empathy, and negotiation benchmarks to previously unseen dialogue domains.

Background

The experiments cover four benchmark domains and demonstrate performance within those settings. However, the authors explicitly state that these benchmarks do not establish generalization to unseen domains and identify broader cross-domain evaluation as unresolved future work.

References

The current benchmarks span persuasion, emotional support, empathy, and negotiation, including cooperative and competitive settings, but they do not establish generalization to unseen dialogue domains; broader cross-domain evaluation remains future work.

— Stop Training Against Soft Users: Mentalized Adversarial Self-Play for Proactive Dialogue  (2610.21156 - Liu et al., 15 Oct 2026) in Appendix, Section “Scope of the Contribution”