ResearchBenchmarksScene & Action Recognition on SUN397 T=30Follow83.73Last Task Accuracy (AT)VILA79.543581.6367583.7385.82325Feb 14, 2026Evaluation ResultsMethodMethodLinksLast Task Accuracy (AT)Average Incremental Accuracy (A_bar)VILABackbone=ViT-B/16Backbone=ViT-B/162026.0283.7389.43