NER-D: leaderboard

German named-entity recognition across thirteen datasets and nine domains. Prompting, in-context examples and trained model results remain distinct, with per-domain F1 scores.

Source: arxiv.org.

Benchmark results

Results are published separately for each benchmark or evaluation setting.

Interactive version: theaggregate.ai/benchmark?slug=ner-d · How It Works · Data refreshed daily, snapshot 2026-11-17.