ResearchBenchmarksMulti-reference Image Generation on MultiBanana (Text and Tone Metrics)Follow3.82Text Quality ScoreDyRef3.5083.5893.673.751Jun 25, 2026Evaluation ResultsMethodMethodLinksText Quality ScoreTone Fidelity ScoreDyRefBase Model=Qwen-2511Base Model=Qwen-25112026.063.823.83Qwen-25112026.063.523.53