Question Answering on ECG-QA (test)
90.9Verification Accuracy (Exact-match)Q-Heart
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Q-HeartModel category=Specialized method2025.10 | 90.9 | — | — | — | — | 60.3 | 32.9 | 61.4 | |
| COUNTSModel category=Our methods, Training stage=Reinforcement Learning, RL steps=30002025.10 | 90.9 | — | — | — | — | 64 | 53.9 | 66.5 | |
| COUNTS (Pre-RL)Model category=Our methods, Training stage=SFT phase, Prompting strategy=few-shot2025.10 | 78.1 | — | — | — | — | 58.9 | 45.2 | 60 | |
| ECG-LMModel category=Specialized method2025.10 | 75.8 | — | — | — | — | 57.4 | 39.9 | 57.7 | |
| M3AEModel category=Specialized method2025.10 | 74.6 | — | — | — | — | 57.1 | 41 | 57.6 | |
| Gemini 2.5 ProModel category=General-purpose LLM2025.10 | 55.7 | — | — | — | — | 30.4 | 17.6 | 30.9 | |
| ChatTsModel category=General-purpose LLM2025.10 | 54.4 | — | — | — | — | 38.2 | 19.1 | 37.2 | |
| ChatTimeModel category=General-purpose LLM2025.10 | 50.1 | — | — | — | — | 30.9 | 15.5 | 30.5 | |
| o4-miniModel category=General-purpose LLM2025.10 | 50 | — | — | — | — | 47.8 | 26.5 | 45.9 | |
| ChatTimetrained on ECG-QA=false, few-shot examples as context=true2025.10 | — | 50.1 | 30.9 | 15.5 | 30.5 | — | — | — | |
| ChatTstrained on ECG-QA=false, few-shot examples as context=true2025.10 | — | 54.4 | 38.2 | 19.1 | 37.2 | — | — | — | |
| ECG-LMtrained on ECG-QA=true2025.10 | — | 75.8 | 57.4 | 39.9 | 57.7 | — | — | — | |
| Gemini 2.5 Protrained on ECG-QA=false, few-shot examples as context=true2025.10 | — | 55.7 | 30.4 | 17.6 | 30.9 | — | — | — | |
| M3AEtrained on ECG-QA=true2025.10 | — | 74.6 | 57.1 | 41 | 57.6 | — | — | — | |
| o4-minitrained on ECG-QA=false, few-shot examples as context=true2025.10 | — | 50 | 47.8 | 26.5 | 45.9 | — | — | — | |
| Q-Hearttrained on ECG-QA=true2025.10 | — | 90.9 | 60.3 | 32.9 | 61.4 | — | — | — | |
| TsLLMtrained on ECG-QA=true2025.10 | — | 85.3 | 60.8 | 49 | 65 | — | — | — |