About SuperLim

SuperLim is a standardized suite for evaluation and analysis of Swedish natural language understanding systems inspired by the GLUE and SuperGLUE benchmarks. The goal of SuperLim and the leaderboard is to help users find the model best suited for their specific task and needs, while also giving model developers the possibility to compare their models with others on a variety of natural language understanding tasks.

The set of benchmark tasks, the initial results, and the leaderboard were co-developed by Språkbanken Text, AISweden, and the National Library of Sweden (KBLab) in a Vinnova-funded two-part project. SuperLim 2.0 extends the test sets developed for SuperLim 1.0 with training sets.

The data can be downloaded directly at SBX, via GitHub, or as a huggingface dataset.