[ { "Model": "CLIP-v1", "distortion": { "human_face": 26.6, "human_limb": 17.2, "object": 34.0, "avg": 19.3 }, "blurry": { "defocused": 50.6, "motion": 63.7, "avg": 56.7 } }, { "Model": "BLIP-v2", "distortion": { "human_face": 3.6, "human_limb": 2.0, "object": 1.1, "avg": 1.9 }, "blurry": { "defocused": 8.3, "motion": 47.2, "avg": 15.0 } }, { "Model": "PickScore-v1", "distortion": { "human_face": 83.4, "human_limb": 68.2, "object": 92.1, "avg": 79.3 }, "blurry": { "defocused": 80.6, "motion": 93.4, "avg": 86.6 } }, { "Model": "HPS-v2.1", "distortion": { "human_face": 60.4, "human_limb": 37.1, "object": 80.3, "avg": 51.7 }, "blurry": { "defocused": 85.7, "motion": 94.6, "avg": 88.6 } }, { "Model": "ImageReward", "distortion": { "human_face": 31.4, "human_limb": 34.4, "object": 40.2, "avg": 33.3 }, "blurry": { "defocused": 77.4, "motion": 86.6, "avg": 82.1 } }, { "Model": "Aesthetics", "distortion": { "human_face": 78.7, "human_limb": 57.1, "object": 51.3, "avg": 52.1 }, "blurry": { "defocused": 90.1, "motion": 93.4, "avg": 91.6 } }, { "Model": "LLaVA-1.5-7b", "distortion": { "human_face": 13.6, "human_limb": 7.3, "object": 9.2, "avg": 10.2 }, "blurry": { "defocused": 7.1, "motion": 19.1, "avg": 13.1 } }, { "Model": "LLaVA-1.5-13b", "distortion": { "human_face": 20.1, "human_limb": 14.6, "object": 13.3, "avg": 16.4 }, "blurry": { "defocused": 18.0, "motion": 34.0, "avg": 26.1 } }, { "Model": "LLaVA-NeXT-7b", "distortion": { "human_face": 28.4, "human_limb": 27.8, "object": 19.0, "avg": 30.1 }, "blurry": { "defocused": 41.7, "motion": 66.1, "avg": 53.9 } }, { "Model": "LLaVA-NeXT-13b", "distortion": { "human_face": 18.9, "human_limb": 27.8, "object": 12.0, "avg": 20.5 }, "blurry": { "defocused": 40.6, "motion": 45.4, "avg": 43.0 } }, { "Model": "Instructblip-7b", "distortion": { "human_face": 12.4, "human_limb": 9.3, "object": 21.0, "avg": 13.3 }, "blurry": { "defocused": 32.3, "motion": 31.1, "avg": 31.7 } }, { "Model": "MiniGPT4-v2", "distortion": { "human_face": 39.6, "human_limb": 39.1, "object": 42.0, "avg": 40.0 }, "blurry": { "defocused": 33.4, "motion": 37.4, "avg": 35.4 } }, { "Model": "Prometheus-Vision-7b", "distortion": { "human_face": 16.6, "human_limb": 17.9, "object": 14.1, "avg": 16.4 }, "blurry": { "defocused": 22.3, "motion": 30.3, "avg": 26.3 } }, { "Model": "Prometheus-Vision-13b", "distortion": { "human_face": 7.1, "human_limb": 4.6, "object": 7.2, "avg": 6.2 }, "blurry": { "defocused": 9.4, "motion": 10.6, "avg": 10.0 } }, { "Model": "Qwen-VL-Chat", "distortion": { "human_face": 14.2, "human_limb": 15.9, "object": 9.4, "avg": 13.6 }, "blurry": { "defocused": 0.9, "motion": 2.1, "avg": 1.4 } }, { "Model": "Internvl-chat-v1-5", "distortion": { "human_face": 97.0, "human_limb": 95.4, "object": 97.1, "avg": 97.1 }, "blurry": { "defocused": 89.7, "motion": 89.7, "avg": 89.7 } }, { "Model": "Idefics2-8b", "distortion": { "human_face": 29.6, "human_limb": 25.8, "object": 2.3, "avg": 21.7 }, "blurry": { "defocused": 70.6, "motion": 46.9, "avg": 58.7 } }, { "Model": "GPT-4-vision", "distortion": { "human_face": 87.6, "human_limb": 57.6, "object": 83.1, "avg": 75.7 }, "blurry": { "defocused": 98.8, "motion": 99.3, "avg": 99.2 } }, { "Model": "GPT-4o", "distortion": { "human_face": 99.4, "human_limb": 78.2, "object": 100.0, "avg": 93.8 }, "blurry": { "defocused": 100.0, "motion": 100.0, "avg": 100.0 } }, { "Model": "Gemini Ultra", "distortion": { "human_face": 73.4, "human_limb": 32.5, "object": 61.0, "avg": 55.7 }, "blurry": { "defocused": 86.5, "motion": 97.3, "avg": 93.9 } }, { "Model": "Claude 3 Opus", "distortion": { "human_face": 26.6, "human_limb": 19.3, "object": 10.7, "avg": 17.6 }, "blurry": { "defocused": 89.6, "motion": 93.3, "avg": 92.7 } } ]