Reproduce “A flexible Bayesian non-parametric mixture model reveals multiple dependencies of swap errors in visual working memory” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Puria Radmard, Paul M. Bays, Máté Lengyel
BNS maps swap probabilities in visual working memory delayed-estimation tasks without fixing in advance how they depend on distractor features. Its flexible model tests cue-feature, report-feature, and joint dependencies. Across datasets, it recovers cue-feature dependence; in direction-cued, location-reported random-dot-motion data, it also finds report-location dependence, consistent with—but not proof of—an encoding contribution, which may also reflect retrieval.
Key insights
- Swap errors in visual working memory can exhibit non-monotonic dependencies on cue features, challenging assumptions of simple proximity-based relationships.
- Errors are not solely driven by retrieval failures; simultaneous and interacting dependencies on both cue and report features suggest a significant role for encoding-stage misbinding errors.
- A flexible, data-driven modeling approach can uncover complex and previously unobserved dependencies in behavioral data, providing new insights into the multifaceted mechanisms of VWM failures.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.