Data foundations for large-scale multimodal clinical models, spanning many data types rather than one.
Audited by MedCheckDai et al., ICML 2025
Key facts
| What it measures | Data foundations for large-scale multimodal clinical models, spanning many data types rather than one. |
| Who should care | Teams pretraining or fine-tuning clinical foundation models. |
| Use it when | Assessing coverage across clinical data modalities. |
| Do not use it for | Ranking off-the-shelf chat models. |
| Score scale | Task-specific metrics across the included datasets. |
| Grader method | unknown |
| First released | 2025 |
| Languages | English |
| Use cases | multimodal, research |
| Citation | Dai et al., ICML 2025 |
| Status | active |
Frequently asked questions
What does CLIMB measure?
Data foundations for large-scale multimodal clinical models, spanning many data types rather than one.
Who should use CLIMB?
Teams pretraining or fine-tuning clinical foundation models. Assessing coverage across clinical data modalities.
What should CLIMB not be used for?
Ranking off-the-shelf chat models.
How are CLIMB scores reported?
Task-specific metrics across the included datasets. Scores from different benchmarks are not comparable to each other.
Is CLIMB independent?
No conflicts of interest have been verified for CLIMB in this registry as of 2026-08-05. Absence of a recorded conflict means none has been verified, not that none exists.
Compare with
Spectrum evaluation for multimodal medical models across many imaging modalities and difficulty levels.
Multimodal evaluation for endoscopy image and video analysis.
Large multimodal benchmark for general medical AI, spanning many imaging modalities, departments and task type
Expert-level medical reasoning questions deliberately built to be harder than licensing exams, with a multimod
The health and medicine subset of a broad expert-level multimodal reasoning benchmark.
Very large medical visual question answering set spanning many modalities and anatomical regions.
