The world’s first large-scale seismic wave model with hundreds of millions of parameters, “Tingting”, was released and will be fully open in 2025

In the field of earthquake monitoring and analysis, a major technological breakthrough was officially released on January 17 at the National Supercomputer Chengdu Center. The third phase test version of the "Tingting" large model, the world's first large-scale seismic wave model with hundreds of millions of parameters, has been launched. This model was jointly developed by the National Supercomputing Center in Chengdu, the Institute of Geophysics of the China Earthquake Administration, and Tsinghua University, marking a leap forward in my country's earthquake research technology.

The purpose of the development of the "True Listen" large model is to improve the recognition accuracy and monitoring capabilities of seismic signals, especially in processing complex seismic waveform data, to provide more efficient and accurate analysis. It is reported that this model has been successfully applied to the processing of 6.8-magnitude earthquake data in the Tingri area of ​​Tibet. Through automated means, 452 pre-earthquake events and 429 aftershock events within 27 hours after the earthquake were identified. This success not only verified the effectiveness of the model, but also provided new technical means for future earthquake monitoring.

Looking forward to the future, the "True Listen" large model will open the fine-tuning and inference framework in 2025, and launch supporting data processing processes. By then, users will be able to use this tool directly on the supercomputing platform for business analysis and scientific research. In the short term, the application of this model will focus on seismic signal identification, seismic activity monitoring, and rapid response to major earthquakes. In the long run, seismology, as an observational science, is inseparable from the in-depth understanding and analysis of observation data.

This development is not only of great significance to scientific researchers, but will also greatly enhance the public's early warning capabilities for earthquake activity. As the model continues to improve, it is expected to provide stronger support for earthquake prediction and disaster prevention and mitigation in the future.