Benchmark
Bridging the Gap MMLU Clinical Knowledge Sesotho (st) human-translated
Multilingual benchmarks compare model performance across languages. Scores are percentages. Higher is better.
- Models and systems measured
- 38
- Scores recorded
- 42
- Sources
- 1
- IQ weight
- 1.20
Score distribution
- 46.4%
- Mean
- 44%
- Median
- 8%
- Std dev
- 37.9%
- 10th percentile
- 54.7%
- 90th percentile
- 30.9%
- Min
This benchmark has a weight of 1.20 for chat models, measured over 6 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model or system | Score |
|---|---|---|
| 1 | GPT 4otable: 2, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 69.4% |
| 2 | GPT-4 Turbotable: 2, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 64.2% |
| 3 | Llama 3 70B Instructtable: 2, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 43.8% |
| 4 | Llama 3 8B Instructtable: 2, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 39.2% |
| 5 | GPT-3.5 Turbotable: 2, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 34.3% |
| 6 | Phi-3 Mini 4K Instructtable: 2, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 30.9% |
Not ranked
The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.
| Model or system | Score |
|---|---|
| Llama 3 70B Instructtable: A.22, row: base, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 43.8% |
| Llama 3 70B Instructtable: A.23, row: base, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 43.8% |
| Llama 3 70B Instructtable: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: High, quantity: 0%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 43.8% |
| Llama 3 70B Instructtable: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: Low, quantity: 0%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 43.8% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: aftable: A.22, training: Winogrande train (small), trainingLanguage: af, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 44.3% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: amtable: A.22, training: Winogrande train (small), trainingLanguage: am, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 42.6% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: bmtable: A.22, training: Winogrande train (small), trainingLanguage: bm, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 33.3% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: entable: A.22, training: Winogrande train (small), trainingLanguage: en, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 40% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: igtable: A.22, training: Winogrande train (small), trainingLanguage: ig, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 43.3% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: nsotable: A.22, training: Winogrande train (small), trainingLanguage: nso, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 40.1% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: sntable: A.22, training: Winogrande train (small), trainingLanguage: sn, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 38.9% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: sttable: A.22, training: Winogrande train (small), trainingLanguage: st, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 40.5% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: tntable: A.22, training: Winogrande train (small), trainingLanguage: tn, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 43% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: tstable: A.22, training: Winogrande train (small), trainingLanguage: ts, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 37.2% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: xhtable: A.22, training: Winogrande train (small), trainingLanguage: xh, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 37.9% |
| Llama 3 70B ITLoRA, dataset: Winogrande train (small), language: zutable: A.22, training: Winogrande train (small), trainingLanguage: zu, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 40.9% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: aftable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: af, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 48.8% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: amtable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: am, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 51.6% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: bmtable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: bm, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 45.8% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: entable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: en, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 43.8% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: igtable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: ig, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 50.9% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: nsotable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: nso, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 54.7% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: sntable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: sn, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 54.7% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: sttable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: st, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 59.2% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: tntable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: tn, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 55.9% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: tstable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: ts, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 54.6% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: xhtable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: xh, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 52.3% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine dev + test + val, language: zutable: A.23, training: MMLU College Medicine dev + test + val, trainingLanguage: zu, task: MMLU Clinical Knowledge, language: st, shots: 5· Bridging the Gap (arXiv:2412.12417v1) data | 52.6% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: High, quantity: 100%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: High, quantity: 100%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 54.7% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: High, quantity: 25%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: High, quantity: 25%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 47.9% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: High, quantity: 50%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: High, quantity: 50%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 50.7% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: High, quantity: 75%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: High, quantity: 75%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 53.3% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: Low, quantity: 100%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: Low, quantity: 100%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 44% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: Low, quantity: 25%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: Low, quantity: 25%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 48.8% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: Low, quantity: 50%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: Low, quantity: 50%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 44.4% |
| Llama 3 70B ITLoRA, dataset: MMLU College Medicine, language: st, quality: Low, quantity: 75%table: A.29, dataset: MMLU College Medicine, trainingLanguage: st, quality: Low, quantity: 75%, task: MMLU Clinical Knowledge· Bridging the Gap (arXiv:2412.12417v1) data | 39.5% |
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)