Clinical Question Answering on MedMCQA
86.1AccuracySAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SAGBackbone=Qwen-4B each, Optimization=w/ GRPO2026.02 | 86.1 | 7 | |
| SAGBackbone=Qwen-4B each, Optimization=w/ CTDE2026.02 | 85.8 | 5.4 | |
| SAGBackbone=Llama-3B each, Optimization=w/ GRPO2026.02 | 85.2 | 6.3 | |
| SAGBackbone=Llama-3B each, Optimization=w/ CTDE2026.02 | 84.7 | 10.2 | |
| SAGBackbone=Qwen-4B each, Optimization=Pre-trained2026.02 | 79.6 | 17.9 | |
| SAGBackbone=Llama-3B each, Optimization=Pre-trained2026.02 | 77.8 | 19.7 | |
| Single giant LLMBackbone=Qwen-72B, Optimization=w/ DPO2026.02 | 75.1 | 15.1 | |
| Single giant LLMBackbone=Llama-70B, Optimization=w/ DPO2026.02 | 74.2 | 30.7 | |
| Single giant LLMBackbone=Qwen-72B, Optimization=w/ PPO2026.02 | 72.5 | 19.3 | |
| Me-LLaMAModel Type=Clinical specialist2026.02 | 71.1 | 26.9 | |
| Single giant LLMBackbone=Llama-70B, Optimization=w/ PPO2026.02 | 70.1 | 18.8 | |
| Single giant LLMBackbone=Qwen-72B, Optimization=Pre-trained2026.02 | 63 | 30.9 | |
| Single giant LLMBackbone=Llama-70B, Optimization=Pre-trained2026.02 | 51.2 | 17.4 | |
| MeditronModel Type=Clinical specialist2026.02 | 48.7 | 28.3 |