ResearchBenchmarksImage-to-interleaved generation on MakeAnything (test)Follow3.15Temporal Coherence (GPT-4o)Loom1.1741.6872.22.713Dec 20, 2025Evaluation ResultsMethodMethodLinksTemporal Coherence (GPT-4o)Temporal Coherence (Human)Reference Faithfulness (GPT-4o)Reference Faithfulness (Human)Semantic Alignment (GPT-4o)Semantic Alignment (Human)Loom2025.123.153.653.854.153.152.95Doubao2025.122.052.152.652.653.553.85Bagelframework=multi-turn d...framework=multi-turn dialogue2025.121.2511.151.55——