NorEval: A Norwegian Language Understanding and Generation Evaluation Benchmark
This provides a standardized tool for evaluating Norwegian LMs, addressing a gap in language-specific benchmarking, though it is incremental as it adapts existing evaluation methods to a new language context.
The paper introduces NorEval, a comprehensive evaluation benchmark for Norwegian language models, covering 24 datasets across understanding and generation tasks for both Bokmål and Nynorsk, and benchmarks 19 models with results integrated into an open framework.
This paper introduces NorEval, a new and comprehensive evaluation suite for large-scale standardized benchmarking of Norwegian generative language models (LMs). NorEval consists of 24 high-quality human-created datasets -- of which five are created from scratch. In contrast to existing benchmarks for Norwegian, NorEval covers a broad spectrum of task categories targeting Norwegian language understanding and generation, establishes human baselines, and focuses on both of the official written standards of the Norwegian language: Bokmål and Nynorsk. All our datasets and a collection of over 100 human-written prompts are integrated into LM Evaluation Harness, ensuring flexible and reproducible evaluation. We describe the NorEval design and present the results of benchmarking 19 open-source pre-trained and instruction-tuned LMs for Norwegian in various scenarios. Our benchmark, evaluation framework, and annotation materials are publicly available.