K-MetBench: A Multi-Dimensional Benchmark for Meteorology Models

K-MetBench is a multi-dimensional benchmark for evaluating meteorology models across accuracy, reasoning quality, geo-cultural alignment, and fine-grained domain coverage. The dataset was created by soyeonbot and was last updated on Hugging Face in April 2026. Its public evaluation protocol uses an explicit advanced benchmark and an explicit reasoning benchmark followed by LLM-as-a-judge evaluation.

Use Cases

• Evaluate model accuracy on fine-grained meteorological concepts based on the described domain coverage.
• Assess the quality of expert reasoning in AI models based on the explicit reasoning benchmark.
• Test geo-cultural alignment of meteorological predictions based on the described locality evaluation dimension.
• Benchmark multimodal AI performance in meteorology based on the described multimodal evaluation dimension.

Strengths

• Designed for multi-dimensional evaluation across accuracy, reasoning, locality, and domain coverage.
• Includes a public evaluation protocol using LLM-as-a-judge for the explicit benchmark splits.
• Last updated on 2026-04-15, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The implicit split of the benchmark is not fully described in the provided metadata.

Provenance

Source

Hugging Face user soyeonbot.

Freshness

Last updated 2026-04-15 18:35:45; freshness should be verified.