8,700 Studies Reviewed. 87.0% Found Biological Effects. The Evidence is Clear.
Shield Your Body
Whole Body / General311 citations

Zhang M, Wang Y, Zou Y, Zhi W, Zhao X, Niu J, Du L, Ma L, Wang L

Bioeffects Seen

Authors not listed · 2025

Share:

This AI benchmarking study has no connection to EMF research and appears misclassified in the database.

Plain English Summary

Summary written for general audiences

This study appears to be about artificial intelligence benchmarking rather than EMF research. The abstract describes 'Humanity's Last Exam,' a new test designed to measure advanced AI capabilities across academic subjects. The study found that current AI models perform poorly on expert-level questions, revealing significant gaps in AI knowledge compared to human experts.

Why This Matters

This study has no relevance to EMF health research. The abstract describes an artificial intelligence benchmark test, not electromagnetic field exposure studies. This appears to be a data entry error or misclassification in the EMF Research Hub database. The reality is that proper scientific databases require accurate categorization to serve researchers and the public effectively. When studies are incorrectly tagged as EMF research, it dilutes the quality of evidence available for understanding electromagnetic field health effects. You deserve access to properly curated research that actually addresses EMF exposure and biological effects.

Exposure Information

Specific exposure levels were not quantified in this study.

Cite This Study
Unknown (2025). Zhang M, Wang Y, Zou Y, Zhi W, Zhao X, Niu J, Du L, Ma L, Wang L.
Show BibTeX
@article{zhang_m_wang_y_zou_y_zhi_w_zhao_x_niu_j_du_l_ma_l_wang_l_ce3584,
  author = {Unknown},
  title = {Zhang M, Wang Y, Zou Y, Zhi W, Zhao X, Niu J, Du L, Ma L, Wang L},
  year = {2025},
  doi = {10.1038/s41586-025-09962-4},
  
}
DOI: 10.1038/s41586-025-09962-4PubMed: 41606155

Quick Questions About This Study

This appears to be a database error. The study examines artificial intelligence capabilities, not electromagnetic field health effects. Proper research databases require accurate categorization to serve users effectively.
It's a 2,500-question test designed to evaluate advanced AI performance across mathematics, humanities, and natural sciences. Current AI models show poor accuracy on these expert-level questions.
No. The study focuses entirely on AI benchmarking and contains no information about electromagnetic fields, radiation exposure, or biological health effects from EMF sources.
The study found that state-of-the-art large language models demonstrate low accuracy on expert-level academic questions, showing significant gaps compared to human knowledge frontiers.
The researchers publicly released Humanity's Last Exam at lastexam.ai for use in AI research and policymaking to better understand current model capabilities and limitations.