ResearchBenchmarksPrompt-side Objective Verification on Expert-annotated prompt-image setFollow97.8AccuracyGem-391.45693.10394.7596.397May 21, 2026Evaluation ResultsMethodMethodLinksAccuracyF1 ScoreGem-3Model=Gem-3Model=Gem-32026.0597.897GPT-5.4Model=GPT-5.4Model=GPT-5.42026.0597.697.2Cl-4.5Model=Cl-4.5Model=Cl-4.52026.0597.498GPT-5.2Model=GPT-5.2Model=GPT-5.22026.0596.695.8GPT-5.4Model=GPT-5.4Model=GPT-5.42026.0595.593.9Gem-3Model=Gem-3Model=Gem-32026.0594.393.3GPT-5.2Model=GPT-5.2Model=GPT-5.22026.0594.293.1Cl-4.5Model=Cl-4.5Model=Cl-4.52026.0591.792.3