ResearchBenchmarksHallucination evaluation on SPD-Faith BenchFollow43.8CRSAGE36.31238.25640.242.144Feb 8, 2026Evaluation ResultsMethodMethodLinksCRDRFSAGEModel=Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.0243.833.7Zoom-RefineModel=Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.0239.731.5VCDModel=Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.0239.431.2SCModel=Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.0238.729.8GreedyModel=Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.0238.229.5SRModel=Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.0237.527.7APIModel=Qwen2.5-VL-7BModel=Qwen2.5-VL-7B2026.0236.628.2