ResearchBenchmarksImage Classification on Spawrious many-to-many hard (test)Follow79.9WGACLIP17.18833.46949.7566.031Jun 30, 2026Evaluation ResultsMethodMethodLinksWGAAccuracyCLIPzero-shot=truezero-shot=true2026.0679.992.8QWEN3-VLzero-shot=truezero-shot=true2026.0673.888.9Chang et al.2026.0666.785.7AutoBackSwap2026.0664.385.7SAM3zero-shot=truezero-shot=true2026.066481.6ERM2026.0651.374.9CORAL2026.0649.874Group DRO2026.0647.171.5AFR2026.0641.368.5DFR2026.0639.167.7ERMaugmentation=heavyaugmentation=heavy2026.0619.645.2