Conflict variants of evidence accumulation models are widely used to explain human cognitive control. However, the generalizability of these models is largely unknown because of the lack of systematic comparisons across multiple conflict tasks and multiple datasets. Here, using a unified amortized Bayesian inference (ABI) framework, we compare the standard DDM and three conflict-specific variants across a benchmark database comprising eight independent studies (N = 1,375). We validate the ABI as a common framework for model comparison, fit DDM, DMC, DSTP, and SSP to each participant, and find that DSTP is the most frequently selected best-fitting model, but its superiority varies more across datasets than across tasks. Exploratory analyses show the 22 parameters cluster into four latent dimensions; the Inhibitory Process factor, of greatest theoretical interest in cognitive control, is unstable. Achieving cross-dataset generality of conflict diffusion models remains a major challenge.
{% callout note %} Click the Cite button above to demo the feature to enable visitors to import publication metadata into their reference management software. {% /callout %}