SEP 8, 2026 · PREPRINT
Limitations of Automated Simulatability: LLM Simulators Can Bypass Explanations
arXiv
A methodological critique identifying systematic vulnerabilities in an automated evaluation protocol, based on qualitative replication and new baselines, but lacking empirical quantification of the magnitude or prevalence of the shortcuts identified.
Study details
DesignQualitative methodological critique and rep…
InterventionClasses-as-concepts baseline and class anon…
ComparatorConSim protocol and original explanation me…