Knowledge-intensive QA on NaturalQuestions (NQ) 5-shot
42.85EM AccuracyADAFUSE (Top-2 Base)
Evaluation results
| Method | Links | |
|---|---|---|
| ADAFUSE (Top-2 Base)Base model selection strategy=Oracle top-2 base selection2026.01 | 42.85 | |
| ADAFUSE (Fixed Base)Base model selection strategy=Fixed base pair (LLaMA-3.1-8B-Instruct and Mistral-7B-Instruct-v0.3)2026.01 | 42.85 | |
| UniTEBase model selection strategy=Fixed base pair (LLaMA-3.1-8B-Instruct and Mistral-7B-Instruct-v0.3)2026.01 | 38.95 | |
| SWEETSPANBase model selection strategy=Fixed base pair (LLaMA-3.1-8B-Instruct and Mistral-7B-Instruct-v0.3)2026.01 | 37.59 | |
| DEEPENBase model selection strategy=Fixed base pair (LLaMA-3.1-8B-Instruct and Mistral-7B-Instruct-v0.3)2026.01 | 37.42 | |
| LLM-BLENDERBase model selection strategy=Fixed base pair (LLaMA-3.1-8B-Instruct and Mistral-7B-Instruct-v0.3)2026.01 | 36.56 | |
| LLaMA-3.1-8B-InstructModel Type=Base Model2026.01 | 34.24 | |
| Mistral-7B-Instruct-v0.3Model Type=Base Model2026.01 | 32.05 | |
| InternLM3-8B-InstructModel Type=Base Model2026.01 | 27.15 | |
| Qwen3-8BModel Type=Base Model2026.01 | 23.85 |