[ {"id": 1, "model": "Waymo Baseline", "year": 2025, "input": "camera", "vision": "-", "language": "-", "action": "-", "rfs_overall": 7.53, "rfs_spotlight": 6.60, "ade_5s": 3.02, "ade_3s": 1.32, "category": "Vision-Action"}, {"id": 2, "model": "Swin-Trajectory", "year": 2025, "input": "camera", "vision": "SwinT", "language": "-", "action": "REG", "rfs_overall": 7.54, "rfs_spotlight": 6.68, "ade_5s": 2.81, "ade_3s": 1.21, "category": "Vision-Action"}, {"id": 3, "model": "DiffusionDrive", "year": 2025, "input": "camera", "vision": "ResNet", "language": "-", "action": "GEN", "rfs_overall": 7.69, "rfs_spotlight": 6.65, "ade_5s": 2.99, "ade_3s": 1.31, "category": "Vision-Action"}, {"id": 4, "model": "RAP-DINO", "year": 2025, "input": "camera", "vision": "DINO", "language": "-", "action": "REG", "rfs_overall": 8.04, "rfs_spotlight": 7.20, "ade_5s": 2.65, "ade_3s": 1.17, "category": "Vision-Action"}, {"id": 5, "model": "OpenEMMA", "year": 2025, "input": "camera, prompt, status", "vision": "Qwen2-VL", "language": "Qwen2-VL", "action": "LH", "rfs_overall": 5.16, "rfs_spotlight": 4.71, "ade_5s": 12.74, "ade_3s": 6.68, "category": "Vision-Language-Action"}, {"id": 6, "model": "HMVLM", "year": 2025, "input": "camera, prompt, command, status", "vision": "ViT", "language": "Qwen2.5-VL", "action": "LH", "rfs_overall": 7.74, "rfs_spotlight": 6.73, "ade_5s": 3.07, "ade_3s": 1.33, "category": "Vision-Language-Action"}, {"id": 7, "model": "AutoVLA", "year": 2025, "input": "camera, prompt, command, status", "vision": "Qwen2.5-VL", "language": "Qwen2.5-VL", "action": "LH", "rfs_overall": 7.56, "rfs_spotlight": 6.94, "ade_5s": 2.96, "ade_3s": 1.35, "category": "Vision-Language-Action"}, {"id": 8, "model": "Poutine", "year": 2025, "input": "camera, prompt, status, command", "vision": "ViT", "language": "Qwen2.5-VL", "action": "LH", "rfs_overall": 7.99, "rfs_spotlight": 6.89, "ade_5s": 2.74, "ade_3s": 1.21, "category": "Vision-Language-Action"}, {"id": 9, "model": "LightEMMA", "year": 2025, "input": "camera, prompt", "vision": "Qwen2.5-VL", "language": "Qwen2.5-VL", "action": "LH", "rfs_overall": 6.52, "rfs_spotlight": 5.71, "ade_5s": 3.73, "ade_3s": 1.71, "category": "Vision-Language-Action"}, {"id": 10, "model": "dVLM-AD", "year": 2025, "input": "camera, prompt, status", "vision": "SigLIP2", "language": "LLaDA-V", "action": "LH", "rfs_overall": 7.63, "rfs_spotlight": null, "ade_5s": 3.02, "ade_3s": 1.29, "category": "Vision-Language-Action"} ]