{"id": "ep_41", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_41/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_41/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"pick empty bottle ab\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.4, \"CFS\": 0.5, \"PP\": 0.3, \"MQ\": 0.5, \"TC\": 0.4, \"VS\": 0.4, \"OS\": 0.4, \"reasoning\": \"The target bottle severely distorts into an unrecognizable blurry shape during the grasp and lift, accompanied by noticeable background warping.\"}"}]}
{"id": "ep_112", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_112/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_112/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put the green marker inside the white bowl\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.6, \"CFS\": 0.7, \"PP\": 0.5, \"MQ\": 0.6, \"TC\": 0.6, \"VS\": 0.65, \"OS\": 0.6, \"reasoning\": \"The green marker severely deforms and loses its shape during manipulation, significantly reducing the visual quality and physical plausibility.\"}"}]}
{"id": "ep_247", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_247/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_247/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Move the small glass object from the right side of the table to the bottom left  corner\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.8, \"CFS\": 0.7, \"PP\": 0.8, \"MQ\": 0.75, \"TC\": 0.7, \"VS\": 0.7, \"OS\": 0.75, \"reasoning\": \"The task is completed, but the predicted video exhibits some temporal flickering and blurring during the object's movement compared to the ground truth.\"}"}]}
{"id": "ep_301", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_301/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_301/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put the orange cup into the black bowl.\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.95, \"CFS\": 0.9, \"PP\": 0.85, \"MQ\": 0.88, \"TC\": 0.85, \"VS\": 0.88, \"OS\": 0.87, \"reasoning\": \"The predicted video successfully completes the task with good motion and temporal coherence, though there is minor morphing of the cup during movement.\"}"}]}
{"id": "ep_86", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_86/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_86/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"little tray storage lemon b\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 1.0, \"CFS\": 0.92, \"PP\": 0.85, \"MQ\": 0.8, \"TC\": 0.88, \"VS\": 0.85, \"OS\": 0.86, \"reasoning\": \"The predicted video successfully completes the task with high visual fidelity, though there is some slight blurring of the robot arm during rapid movement compared to the ground truth.\"}"}]}
{"id": "ep_340", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_340/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_340/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Open the machine\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.9, \"CFS\": 0.85, \"PP\": 0.85, \"MQ\": 0.8, \"TC\": 0.85, \"VS\": 0.8, \"OS\": 0.83, \"reasoning\": \"The predicted video successfully visualizes the robot opening the machine with good temporal consistency, though there are minor visual differences from the ground truth execution.\"}"}]}
{"id": "ep_2", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_2/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_2/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"pick apple b\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 1.0, \"CFS\": 0.85, \"PP\": 0.85, \"MQ\": 0.8, \"TC\": 0.9, \"VS\": 0.82, \"OS\": 0.84, \"reasoning\": \"The predicted video successfully completes the task with high visual fidelity, though the robot's gripper motion is slightly less smooth than the ground truth.\"}"}]}
{"id": "ep_398", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_398/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_398/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put the yellow block in the box.\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.95, \"CFS\": 0.85, \"PP\": 0.8, \"MQ\": 0.85, \"TC\": 0.8, \"VS\": 0.85, \"OS\": 0.85, \"reasoning\": \"The predicted video correctly and smoothly completes the task, matching the ground truth well, though minor fuzziness and blurring around the gripper slightly reduce temporal consistency.\"}"}]}
{"id": "ep_116", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_116/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_116/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Slightly open the wooden box by sliding the lid off\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.95, \"CFS\": 0.85, \"PP\": 0.8, \"MQ\": 0.85, \"TC\": 0.8, \"VS\": 0.85, \"OS\": 0.85, \"reasoning\": \"The predicted video correctly shows the robot sliding the lid off the wooden box, though there are minor temporal blurring artifacts during the motion.\"}"}]}
{"id": "ep_278", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_278/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_278/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put some sweets in the bowl\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.8, \"CFS\": 0.85, \"PP\": 0.8, \"MQ\": 0.85, \"TC\": 0.8, \"VS\": 0.85, \"OS\": 0.82, \"reasoning\": \"The predicted video successfully completes the task, but the small target objects exhibit slight blurring during the grasping and placing motion.\"}"}]}
{"id": "ep_319", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_319/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_319/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Move the white and black ball to the left\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.5, \"CFS\": 0.6, \"PP\": 0.3, \"MQ\": 0.4, \"TC\": 0.4, \"VS\": 0.5, \"OS\": 0.4, \"reasoning\": \"The predicted video exhibits severe object morphing and blurring when the gripper interacts with the ball, reducing physical plausibility and temporal consistency.\"}"}]}
{"id": "ep_359", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_359/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_359/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put all of the yellow things in the cup\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.8, \"CFS\": 0.75, \"PP\": 0.75, \"MQ\": 0.8, \"TC\": 0.85, \"VS\": 0.75, \"OS\": 0.78, \"reasoning\": \"The robot successfully places the yellow objects into the cup, but the predicted video exhibits minor blurring and slight temporal flickering during the grasping actions.\"}"}]}
{"id": "ep_327", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_327/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_327/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Use the napkin to wipe the counter.\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.8, \"CFS\": 0.75, \"PP\": 0.75, \"MQ\": 0.75, \"TC\": 0.7, \"VS\": 0.72, \"OS\": 0.73, \"reasoning\": \"The robot successfully completes the wiping task, but the predicted video suffers from noticeable blurring and structural distortion of the robot arm and napkin in the later frames.\"}"}]}
{"id": "ep_26", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_26/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_26/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"pick up the lemon at\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 1.0, \"CFS\": 0.85, \"PP\": 0.85, \"MQ\": 0.8, \"TC\": 0.85, \"VS\": 0.82, \"OS\": 0.84, \"reasoning\": \"The predicted video successfully completes the task with high visual fidelity, though there is some slight blurring of the gripper and lemon during the lifting motion compared to the ground truth.\"}"}]}
{"id": "ep_310", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_310/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_310/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Pick up the green marker from the table and put it in the grey bowl\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 1.0, \"OPS\": 0.9, \"CFS\": 0.85, \"PP\": 0.85, \"MQ\": 0.82, \"TC\": 0.8, \"VS\": 0.84, \"OS\": 0.83, \"reasoning\": \"The predicted video completes the task successfully but shows minor temporal flickering and slight blurriness around the marker during the interaction.\"}"}]}
{"id": "ep_223", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_223/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_223/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Turn on the switch on the adapter\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.6, \"CFS\": 0.5, \"PP\": 0.4, \"MQ\": 0.4, \"TC\": 0.4, \"VS\": 0.45, \"OS\": 0.4, \"reasoning\": \"Severe blurring and morphological distortion of the robot arm and target object during the interaction phase prevent clear verification of task completion.\"}"}]}
{"id": "ep_250", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_250/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_250/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put the black masking tape into the box.\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.6, \"CFS\": 0.5, \"PP\": 0.4, \"MQ\": 0.5, \"TC\": 0.45, \"VS\": 0.5, \"OS\": 0.45, \"reasoning\": \"The predicted video suffers from severe object morphing and temporal inconsistencies, making it unclear if the tape successfully enters the box.\"}"}]}
{"id": "ep_190", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_190/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_190/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put the black clothing on the table\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.3, \"CFS\": 0.5, \"PP\": 0.2, \"MQ\": 0.5, \"TC\": 0.3, \"VS\": 0.4, \"OS\": 0.3, \"reasoning\": \"The black clothing completely vanishes from the robot's gripper early in the predicted video, violating physics and preventing task completion.\"}"}]}
{"id": "ep_154", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_154/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_154/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Move the bottle from the table to the tray\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.8, \"CFS\": 0.5, \"PP\": 0.4, \"MQ\": 0.5, \"TC\": 0.6, \"VS\": 0.5, \"OS\": 0.4, \"reasoning\": \"The robot in the predicted video fails to grasp and lift the bottle, completely missing the task objective and exhibiting poor physical interaction.\"}"}]}
{"id": "ep_198", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_198/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_198/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Pick up the light blue toy from the table and put it in the green bowl\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.5, \"CFS\": 0.6, \"PP\": 0.3, \"MQ\": 0.4, \"TC\": 0.3, \"VS\": 0.45, \"OS\": 0.35, \"reasoning\": \"The predicted video suffers from severe morphing and flickering during the grasping phase, causing the target object to vanish rather than being visibly placed into the bowl.\"}"}]}
{"id": "ep_137", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_137/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_137/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put the cloth on the black rod\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.2, \"CFS\": 0.2, \"PP\": 0.1, \"MQ\": 0.2, \"TC\": 0.2, \"VS\": 0.2, \"OS\": 0.2, \"reasoning\": \"The predicted video suffers from severe visual distortion, blurring, and loss of object structure shortly after motion begins, leading to task failure.\"}"}]}
{"id": "ep_338", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_338/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_338/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Pick up the blue block, put it in the drawer and close the drawer.\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.5, \"CFS\": 0.4, \"PP\": 0.3, \"MQ\": 0.4, \"TC\": 0.35, \"VS\": 0.4, \"OS\": 0.35, \"reasoning\": \"The predicted video suffers from severe morphing and temporal inconsistencies, causing the blue block to lose its shape and the task (placing the block and closing the drawer) to remain uncompleted.\"}"}]}
{"id": "ep_100", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_100/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_100/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Put the towel on the table\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.5, \"CFS\": 0.3, \"PP\": 0.2, \"MQ\": 0.2, \"TC\": 0.3, \"VS\": 0.3, \"OS\": 0.2, \"reasoning\": \"The predicted video suffers from severe blurring and spatial distortion, failing to depict the robot grasping or moving the towel at all.\"}"}]}
{"id": "ep_342", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_342/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_342/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Press one of the buttons on the coffee making machine\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.0, \"CFS\": 0.0, \"PP\": 0.0, \"MQ\": 0.0, \"TC\": 0.0, \"VS\": 0.0, \"OS\": 0.0, \"reasoning\": \"Since no video inputs were actually provided in the prompt, metrics are set to minimum default values.\"}"}]}
{"id": "ep_96", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_96/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_96/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"pick cup a\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.5, \"CFS\": 0.6, \"PP\": 0.2, \"MQ\": 0.4, \"TC\": 0.4, \"VS\": 0.5, \"OS\": 0.3, \"reasoning\": \"The predicted video fails to complete the task because the target cup disappears or melts into the gripper upon contact, severely violating physical plausibility.\"}"}]}
{"id": "ep_197", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_197/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_197/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Pick up the saltcellar from the pot and put it on the table\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.7, \"CFS\": 0.6, \"PP\": 0.65, \"MQ\": 0.7, \"TC\": 0.6, \"VS\": 0.65, \"OS\": 0.62, \"reasoning\": \"The predicted video exhibits temporal inconsistency and the object is dropped before reaching the table, leading to a failed task completion.\"}"}]}
{"id": "ep_185", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_185/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_185/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Slide the mug towards you\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.4, \"CFS\": 0.5, \"PP\": 0.1, \"MQ\": 0.4, \"TC\": 0.3, \"VS\": 0.5, \"OS\": 0.3, \"reasoning\": \"The target mug completely disappears and morphs into the robot's gripper upon interaction, resulting in a failure to complete the task and severely violating physical plausibility.\"}"}]}
{"id": "ep_81", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_81/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_81/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"place apple c\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.2, \"CFS\": 0.4, \"PP\": 0.2, \"MQ\": 0.5, \"TC\": 0.4, \"VS\": 0.4, \"OS\": 0.3, \"reasoning\": \"The target apple disappears completely when grasped by the robot, leading to a failure in completing the placement task.\"}"}]}
{"id": "ep_158", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_158/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_158/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"Remove the clamp from the bowl and put it in the clear bowl with white decorating tape\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.5, \"CFS\": 0.5, \"PP\": 0.4, \"MQ\": 0.45, \"TC\": 0.4, \"VS\": 0.5, \"OS\": 0.45, \"reasoning\": \"The predicted video suffers from severe blurring and temporal inconsistency during the object interaction, causing the clamp to become unrecognizable and failing to complete the task.\"}"}]}
{"id": "ep_94", "pred_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_94/pred_exterior.mp4", "gt_video": "/mnt/users/zirui/mizirui_benchmark/World_Fufu/eval_set/episode_94/gt_exterior.mp4", "conversations": [{"from": "human", "value": "<video><video>\nYou are an expert evaluator for robot manipulation world model videos.\nYou receive two videos: the FIRST is the PREDICTED video (world model output), the SECOND is the GROUND TRUTH video (real robot execution).\nTask instruction: \"pick apple b\"\n\nEvaluate the predicted video and respond ONLY with valid JSON containing these fields:\nTCR (0 or 1), OPS (0-1), CFS (0-1), PP (0-100), MQ (0-100), TC (0-100), VS (0-100), OS (0-100), reasoning (one sentence)."}, {"from": "gpt", "value": "{\"TCR\": 0.0, \"OPS\": 0.5, \"CFS\": 0.4, \"PP\": 0.3, \"MQ\": 0.4, \"TC\": 0.4, \"VS\": 0.4, \"OS\": 0.35, \"reasoning\": \"The predicted video fails to complete the task as the robot's gripper completely disappears and the target object becomes severely distorted during the attempted grasp.\"}"}]}