MI-CXR - Interval-wise Change Reasoning: leaderboard
Interval-wise Change Reasoning questions of MI-CXR (2026), five-option questions over five-visit chest X-ray timelines from MIMIC-CXR, answered zero-shot from all five images: find the interval that matches a described change and interpret it. Accuracy in percent (chance 20); higher is better.
Metric: Interval-wise Change Reasoning: find the interval matching a described change, with the interval not given, and interpret it, five-option multiple-choice accuracy (%, times 100; chance 20) on MI-CXR, longitudinal reasoning over five-visit chest X-ray timelines from MIMIC-CXR (5,311 questions), zero-shot single-step prompting with all five images, temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around May 2028. 14 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3.5-38B | 57.1 | #1026 |
| 2 | InternVL3.5-8B | 55.2 | #1618 |
| 3 | Gemini 3 Pro (Preview) | 45.7 | #123 |
| 4 | Claude Sonnet 4.5 | 44.2 | #267 |
| 5 | GPT-5.2 | 43.8 | #190 |
| 6 | MedGemma-27B-IT | 42.9 | #1359 |
| 7 | InternVL3.5-14B | 29.3 | #1519 |
| 8 | MedGemma-4B | 28.1 | #2579 |
| 9 | IDEFICS2-8B | 24.6 | #5025 |
| 10 | Qwen 3 VL 32B | 22.4 | #624 |
| 11 | Lingshu-7B | 18.9 | #1646 |
| 12 | MI-CXR DeepSeek-VL-16B (checkpoint unspecified) | 18.6 | |
| 13 | Lingshu-32B | 16.7 | #1125 |
| 14 | Qwen 3 VL 8B | 16.4 | #927 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=mi-cxr-interval-wise-change-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-22.