SeisMark: A Large-Scale Open Benchmark for Robust 3D Seismic Fault Detection

We introduce SeisMark, a large-scale open benchmark designed to bridge the gap between verifiable ground truth and realistic texture in 3D seismic fault detection. Using a novel pipeline merging procedural geology with diffusion-based synthesis, we produce domain-realistic (survey-specific) textured volumes that expose significant brittleness in existing models masked by simplified physics data. Experiments demonstrate that SeisMark acts as a rigorous discriminator, distinguishing robust modern architecture from legacy model that suffers performance collapse under realistic domain shifts. We release this benchmark to the community to serve as a verifiable standard for developing trustworthy, deployment-ready AI for safety-critical subsurface applications.

AI models are increasingly used to detect underground faults in 3D seismic data, but evaluating their real-world performance is difficult. Existing datasets rely on overly simplified simulations, which hides where these models fail when applied to real field data. To bridge the gap between verifiable ground truth and realistic textures, we introduce SeisMark, a new large-scale open benchmark. By combining procedural geology with advanced diffusion-based generative models, we synthesize highly realistic 3D test data. Our experiments reveal that SeisMark effectively exposes legacy models whose accuracy collapses on the realistic textures, while identifying robust architectures that truly generalize. We release this benchmark to the community to accelerate the development of trustworthy, deployment-ready AI for safety-critical subsurface applications.