The TMT Benchmark uses the Stanford Topic Modeling Toolbox (TMT) to learn a topic model using Latent Dirichlet Allocation (LDA). This task is based on an example provided by TMT’s authors. Each workload of the TMT Benchmark performs a different number of learning iterations.


The TMT Benchmark is externally single‐threaded and internally multi‐threaded. It creates a large number of threads, each of which is only very short‐lived.


We are grateful to Daniel Ramage, one of the authors of TMT, for his support.