ResearchBenchmarksText-to-Image Generation on Human Evaluation (In-Distribution)Follow85Win Ratio (Overall)SEER76.6878.848183.16Jan 28, 2026Evaluation ResultsMethodMethodLinksWin Ratio (Overall)Win Ratio (Simple)Win Ratio (Hard)SEEROpponent Model=Base (CoT)Opponent Model=Base (CoT)2026.01857890SEEROpponent Model=Show-o2Opponent Model=Show-o22026.01857889SEEROpponent Model=Blip3-oOpponent Model=Blip3-o2026.01807683SEEROpponent Model=BagelOpponent Model=Bagel2026.01776486SEEROpponent Model=Bagel-T...Opponent Model=Bagel-Think2026.01776386