Figure/Table detail

ST-Crime: A Retrieval Augmented Pre-trained Foundation Model for Environment-Dependent Crime Spatio-Temporal Prediction
WANG Tao, ZHANG Yifan, CHEN Peng
Journal of Geo-information Science, 2026, 28(1): 209-221.   DOI: 10.12082/dqxxkx.2026.250448

推理 评估指标 LR LSTM MiST CF AGL-STAN UrbanGPT UniFlow ST-Crime
少样本推理 Macro-F1 0.323 1 0.438 3 0.419 5 0.446 7 0.484 1 0.596 3 0.615 4 0.658 6
Micro-F1 0.285 5 0.372 0 0.358 8 0.358 4 0.413 7 0.528 1 0.577 2 0.596 9
零样本推理 Macro-F1 0.287 5 0.406 0 0.403 7 0.411 7 0.438 6 0.520 1 0.559 8 0.603 1
Micro-F1 0.271 8 0.361 4 0.349 1 0.372 0 0.396 1 0.473 4 0.516 2 0.565 3
Tab. 5 Few-shot and zero-shot inference performance
Other figure/table from this article