Reproduce “DES-JEPA: Learning a Shared Vocabulary for High-Dimensional Discrete Event Sequences” 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
Hugo Math, Maximilian Forstenhäusler, Christos Anagnostopoulos
DES-JEPA and DES-PFN enable zero-shot next-event prediction across unaligned event vocabularies. They learn predictive codes from synthetic streams and use context examples to map predictions back to local labels. On BGL with 120 event types, DES-PFN reaches top-1 accuracy of 0.136±0.020, compared with 0.063 for fully trained SAHP. The comparison uses different training conditions and is not controlled.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.