Predict any LLM's score on any benchmark.
Use the code to reproduce the paper, or download the score matrix behind the predictor.
Report benchmark scores for a model. Include the model, benchmark, score, evaluation setting, effort, and source; we will review provenance before adding it to the matrix.
Link preview
GitHub - microsoft/benchpress: BenchPress: calibrated LLM benchmark score completion
BenchPress: calibrated LLM benchmark score completion - microsoft/benchpress GitHub · github.com
Comments