MM-Health: 34,746 Multimodal News Articles for Health Misinformation Detection

MM-Health is a large-scale multimodal dataset containing 34,746 news articles with both textual and visual information. It is designed for benchmarking the detection of health misinformation, including content generated by both humans and AI. The dataset was created by author zzha6204 and was last updated on huggingface in November 2025.

Use Cases

• Train multimodal classifiers to detect health misinformation based on combined text and image content.
• Benchmark AI-generated content detection models against human-written misinformation.
• Analyze the characteristics and prevalence of health misinformation across a large corpus of news articles.

Strengths

• Contains 34,746 news articles, making it a large-scale resource.
• Integrates both textual and visual information, enabling multimodal analysis.
• Specifically designed for benchmarking both human and AI-generated health misinformation detection.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but specific file formats, sample data, and license information are unavailable.
• Data may reflect temporal and source bias inherent to the news articles collected.

Provenance

Source

huggingface, author zzha6204

Collection Method

Likely collected and annotated from online news sources, as described in the associated GitHub repository.

Freshness

Last updated 2025-11-20 04:36:36; freshness should be verified.