Question Answering on MMLU-CK (test)
60.95AccuracyPubMed Reasoner
Evaluation Results
| Method | Links | |
|---|---|---|
| PubMed ReasonerBackbone=Qwen2.5-72B2026.03 | 60.95 | |
| Self-reflection agentBackbone=Qwen2.5-72B2026.03 | 59.67 | |
| Without retrievalBackbone=Qwen2.5-72B2026.03 | 58.36 | |
| Without retrievalBackbone=Qwen2.5-7B2026.03 | 56.64 | |
| Self-reflection agentBackbone=Qwen2.5-7B2026.03 | 56.11 | |
| PubMed ReasonerBackbone=Qwen2.5-7B2026.03 | 56.09 | |
| Without retrievalBackbone=Qwen2.5-1.5B2026.03 | 55.67 | |
| Self-reflection agentBackbone=Qwen2.5-1.5B2026.03 | 51.44 | |
| PubMed ReasonerBackbone=Qwen2.5-1.5B2026.03 | 51.42 |