Scoping review synthesizes NLP evaluation methodology concerns into taxonomy with practical checklist

Researchers conducted a scoping review of evaluation concerns in natural language processing (NLP) and developed a taxonomy that synthesizes recurring positions and trade-offs within the field.

The work situates contemporary debates about large language model (LLM) evaluation within their historical context in NLP, a field with a long history of methodological reflection on evaluation practices.

The taxonomy includes practical implications, including a structured checklist designed to support more deliberate evaluation design and interpretation.