ResearchBenchmarksEvaluator Preference Collapse Analysis on Multimodal EPC Text Phase 1Follow1.348PCIGPT-4o1.28061.31431.3481.3817Jun 15, 2026Evaluation ResultsMethodMethodLinksPCIComparison vs RandomComparison vs Self-EvalGPT-4oEvaluator=GPT-4oEvaluator=GPT-4o2026.061.348—2.92