CaReCoS - Cough Count: leaderboard

Metric: Accuracy (%) of one-to-two-sentence answers to clinically grounded questions about a mel-spectrogram image of the recording, judged by an LLM against the reference (strict factual matching for explicit questions, relaxed semantic matching for inferred ones); the 168 questions on EPFL multimodal cough detection recordings about temporal cough features (cough count, per-cough duration, inter-cough intervals); higher is better. Source: arxiv.org. Saturation forecast: Around December 2026. 9 models tracked.

Top models

#ModelScore
1GPT-5.146.4
2Gemini 2.5 Pro38.7
3Gemini 2.5 Flash38.1
4Qwen2.5-Omni-7B13.1

Interactive version: theaggregate.ai/benchmark?slug=carecos-cough-count · How It Works · Data refreshed daily, snapshot 2026-09-29.