ECG-Reasoning-Benchmark (MIMIC-IV-ECG) - Diagnosis Given Reference Reasoning: leaderboard
Metric: Diagnosis accuracy (%) at the final decision step when the complete ground-truth reasoning history is supplied in the prompt (GT-RDA), on the 3,355 MIMIC-IV-ECG samples, multi-turn ECG reasoning chains for 17 core diagnoses (balanced positive and negative cases), 12-lead ECGs given as plotted images (OpenTSLM as a 100 Hz time series), temperature 0; each step's answer is checked for semantic agreement with the ground truth by a Gemini-3-Flash judge; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash (Preview) | 91.25 | #78 |
| 2 | Gemini 2.5 Flash | 89.73 | #237 |
| 3 | Llama 3.2 90B Vision Instruct | 88.39 | #611 |
| 4 | Gemini 2.5 Pro | 88.06 | #145 |
| 5 | Qwen 3 VL 32B Instruct | 88 | #276 |
| 6 | Qwen 3 VL 8B Instruct | 86.19 | #401 |
| 7 | GPT-5.2 | 83.83 | #105 |
| 8 | GPT-5 Mini (2025-08-07) | 75.65 | #165 |
| 9 | Llama 3.2 11B Instruct | 74.31 | #1112 |
| 10 | MedGemma-4B-IT | 30.72 | #842 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=ecg-reasoning-benchmark-mimic-iv-ecg-diagnosis-given-reference-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-11.