Skip to content

Benchmark

FormosanBench summarization ROUGE-L (Amis, 10-shot)

Multilingual benchmarks compare model performance across languages. Scores are percentages. Higher is better.

Models measured
3
Scores recorded
3
Sources
1
IQ weight
No weight published

Score distribution

14.9%
Mean
7.7%
Median
10.7%
Std dev
7.2%
10th percentile
30%
90th percentile
7.2%
Min
7.2%30%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1GPT 4o10-shot, 10 in-context examples, Mandarin prompt· FormosanBench (arXiv:2506.21563v1) data30%
2mistral-7b-v0.310-shot, 10 in-context examples, Mandarin prompt· FormosanBench (arXiv:2506.21563v1) data7.7%
3Llama 3.1 8B10-shot, 10 in-context examples, Mandarin prompt· FormosanBench (arXiv:2506.21563v1) data7.2%

Data credits: Lin, Chen and Zhang, FormosanBench: Benchmarking Low-Resource Austronesian Languages in the Era of Large Language Models, arXiv:2506.21563v1, CC BY 4.0. Scores extracted from Tables 2-4. (licence)