Benchmark
Bridging the Gap Winogrande Xhosa (xh) human translation back-translated to English
Multilingual benchmarks compare model performance across languages. Scores are percentages. Higher is better.
- Models measured
- 6
- Scores recorded
- 6
- Sources
- 1
- IQ weight
- No weight published
Score distribution
- 58.4%
- Mean
- 57.2%
- Median
- 6.3%
- Std dev
- 54.1%
- 10th percentile
- 68.4%
- 90th percentile
- 51%
- Min
Results
The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.
| Model | Score |
|---|---|
| GPT 4otable: A.11, section: back-translation-accuracy, task: Winogrande, language: xh, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 68.4% |
| GPT-3.5 Turbotable: A.11, section: back-translation-accuracy, task: Winogrande, language: xh, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 54.1% |
| GPT-4 Turbotable: A.11, section: back-translation-accuracy, task: Winogrande, language: xh, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 65.3% |
| Llama 3 70B Instructtable: A.11, section: back-translation-accuracy, task: Winogrande, language: xh, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 54.5% |
| Llama 3 8B Instructtable: A.11, section: back-translation-accuracy, task: Winogrande, language: xh, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 51% |
| Phi-3 Mini 4K Instructtable: A.11, section: back-translation-accuracy, task: Winogrande, language: xh, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 57.2% |
Data credits: Tuka Alhanai, Adam Kasumovic, Mohammad Ghassemi, Aven Zitzelberger, Jessica Lundin and Guillaume Chabot-Couture. Bridging the Gap: Enhancing LLM Performance for Low-Resource African Languages with New Benchmarks, Fine-Tuning, and Cultural Adjustments, arXiv:2412.12417v1. CC BY 4.0 article numeric/protocol extract, restructured with values unchanged. No task or model rights asserted. (licence)