The SteuerEx Benchmark is an evaluation framework for assessing language model performance on German tax law questions. This benchmark uses a rigorous bootstrap-based evaluation methodology to ensure reliable and reproducible results. This benchmark is based on the SteuerLLM paper (SteuerLLM: Local specialized large language model for German tax law analysis).
Official Server: https://steuerllm.i5.ai.fau.de/benchmark
SteuerLLM Demo:… See the full description on the dataset page: https://huggingface.co/datasets/windprak/SteuerEx.