NER-D - CO-Fun - 0-Shot - Fuzzy Macro F1: leaderboard

German13datasets; per-columnAverageexcludesSwissNERforallcolumns. SwissNER0shotonly. Exactentitystring+typematch; fuzzytoken-set-ratiothreshold.90. Microversusmacroretained. ExplicitHFcheckpoints:Ministral8BInstruct2410, Qwen3.5-9B, Gemma4-12B-it. GPT5.4mini0shotonly. Per-datasetLoRAadaptersnotattributedtobase; checkpointreleaseunresolved. Metric: F1 (0–1). Source: https://arxiv.org/html/2611.18257v1

Metric: F1 (0–1). Source: arxiv.org. Saturation forecast: Estimated already saturated. 4 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Ministral-8B-Instruct-24100.844#4569
2Gemma 4 12B (IT)0.805#733
3Qwen 3.5 9B0.783#822
4GPT-5.4 Mini0.774#401

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=ner-d-co-fun-0-shot-fuzzy-macro-f1 · How It Works · Data refreshed daily, snapshot 2026-11-17.