MI-CXR - Single Event Localization: leaderboard

Single Event Localization 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: identify the interval in which one abnormality appears or resolves. Accuracy in percent (chance 20); higher is better.

Metric: Temporal Event Localization, single emergence or resolution: identify the interval in which one abnormality appears or resolves, 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 2029. 14 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.233.4#185
2InternVL3.5-38B29.8#1021
3Qwen 3 VL 32B25.8#622
4InternVL3.5-14B24.8#1504
5Gemini 3 Pro (Preview)24.6#119
6InternVL3.5-8B23.9#1604
7Qwen 3 VL 8B23.7#917
8Lingshu-7B23#1635
9Claude Sonnet 4.522.6#261
10MI-CXR DeepSeek-VL-16B (checkpoint unspecified)22.3
11Lingshu-32B22.1#1113
12MedGemma-27B-IT21.5#1350
13MedGemma-4B17.4#2557
14IDEFICS2-8B16.5#5013

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=mi-cxr-single-event-localization · How It Works · Data refreshed daily, snapshot 2026-11-21.