paulpacaud/ur5fail_val_dataset

Guardian Failure Detection Dataset

This dataset is part of the Guardian project: Detecting Robotic Planning and Execution Errors with Vision-Language Models. It contains annotated robotic manipulation failure data for training and evaluating Vision-Language Models (VLMs) on failure detection tasks.

Guardian introduces an automated failure generation approach that procedurally perturbs successful robot trajectories to produce diverse planning failures and execution failures, each annotated with fine-grained failure categories and step-by-step reasoning traces.

Dataset Family

This repository is one split of the Guardian dataset family. The full collection includes:

Dataset	Source	Execution (Train / Val / Test)	Planning (Train / Val / Test)
RLBench-Fail	RLBench simulator (52 tasks)	12,358 / 1,000 / 1,000	5,808 / 500 / 500
BridgeDataV2-Fail	BridgeDataV2 real-robot data	7,830 / 1,000 / 1,000	4,880 / 500 / 500
UR5-Fail	UR5 robot, 3 cameras, 34 tasks	400 / 30 / 140	200 / 30 / 140
RoboFail	Existing benchmark (test only)	153	30

Directory Structure

Each dataset split is organized as follows:

<dataset_name>/
├── metadata_execution.jsonl             # Rich metadata for execution verification samples
├── metadata_planning.jsonl              # Rich metadata for planning verification samples
├── internVL_dataset_execution_vanilla.jsonl   # InternVL fine-tuning format (execution, no CoT)
├── internVL_dataset_execution_thinking.jsonl  # InternVL fine-tuning format (execution, with CoT)
├── internVL_dataset_planning_vanilla.jsonl    # InternVL fine-tuning format (planning, no CoT)
├── internVL_dataset_planning_thinking.jsonl   # InternVL fine-tuning format (planning, with CoT)
└── records/                             # Images organized by task, failure mode, and episode
    └── <taskvar>/
        └── <failure_mode>/
            └── ep_<id>/
                └── <subtask_id>/
                    ├── start_img_viewpoint_0.png
                    ├── start_img_viewpoint_1.png
                    ├── start_img_viewpoint_2.png
                    ├── end_img_viewpoint_0.png
                    ├── end_img_viewpoint_1.png
                    └── end_img_viewpoint_2.png

Task Types

Execution Verification

Given a high-level task goal, a subtask description, and multi-view images before and after the subtask execution, the model must determine whether the subtask was completed successfully and categorize the failure mode.

• Input: 6 images (3 viewpoints x 2 timesteps: start and end) + task instruction + subtask description
• Output: Boolean (success/failure) + failure category

Planning Verification

Given a high-level task goal, a proposed plan, and the initial scene image, the model must determine whether the plan is correct and categorize the failure mode.

• Input: 1 image (front view of the initial scene) + task instruction + proposed plan
• Output: Boolean (correct/incorrect) + failure category

Failure Categories

Execution Failures

Category	Description
success	The subtask was completed successfully
no gripper close	The gripper is correctly positioned but did not close its jaws
imprecise grasping/pushing	The gripper attempted the correct object but missed due to inaccurate positioning
wrong object state or placement	The correct object was manipulated but the final state or placement is wrong
wrong object manipulated	The gripper manipulated the wrong object
no progress	Neither the scene state nor the robot's configuration changed in any meaningful way

Planning Failures

Category	Description
success	The plan is correct
missing subtasks	One or several required subtasks are missing from the plan
wrong object manipulated	One or several subtasks manipulate the wrong object
wrong object state or placement	One or several subtasks select the wrong target, location, or state
wrong order	Subtasks are not in the right order, breaking causal dependencies
contradictory subtasks	Some subtasks conflict with each other

Data Formats

Metadata Files (metadata_*.jsonl)

These files contain rich per-sample annotations. Each line is a JSON object.

Execution metadata (metadata_execution.jsonl):

{
  "taskvar": "real_hang_mug+1",
  "task_instruction": "take the pink mug and put it on the middle part of the hanger",
  "episode_id": 0,
  "images": [
    "records/real_hang_mug+1/translation_object/ep_0/1/start_img_viewpoint_0.png",
    "records/real_hang_mug+1/translation_object/ep_0/1/start_img_viewpoint_1.png",
    "records/real_hang_mug+1/translation_object/ep_0/1/start_img_viewpoint_2.png",
    "records/real_hang_mug+1/translation_object/ep_0/1/end_img_viewpoint_0.png",
    "records/real_hang_mug+1/translation_object/ep_0/1/end_img_viewpoint_1.png",
    "records/real_hang_mug+1/translation_object/ep_0/1/end_img_viewpoint_2.png"
  ],
  "failure_mode": "translation_object",
  "detailed_subtask_name": "move grasped object to hanger",
  "failure_reason": "the gripper remains closed from start to end...",
  "visible_objects": ["pink mug", "blue mug", "green mug", "wooden mug hanger", "navy table top", "robot arm equipped with a gripper"],
  "plan": ["grasp pink mug", "move grasped object to hanger", "release"],
  "reward": 0,
  "planning_reward": 1,
  "execution_reward": 0
}

Planning metadata (metadata_planning.jsonl):

{
  "taskvar": "real_hang_mug+1",
  "episode_id": 4,
  "task_instruction": "put into the right part of the wooden hanger the mug that is the closest to the wooden hanger",
  "plan": ["grasp pink mug", "release pink mug", "move grasped pink mug to right part of hanger", "release pink mug"],
  "images": [
    "records/real_hang_mug+1/translation_target/ep_4/0/start_img_viewpoint_0.png"
  ],
  "planning_reward": 0,
  "execution_reward": 1,
  "failure_reason": "the plan introduces a contradiction by grasping then immediately releasing",
  "detailed_subtask_name": null,
  "failure_mode": "contradictory subtasks",
  "visible_objects": ["green mug", "pink mug", "blue mug", "wooden cup hanger", "navy table top", "robot arm equipped with a gripper"],
  "old_visible_objects": [],
  "old_task_instruction": "take the pink mug and put it on the middle part of the hanger",
  "old_plan": ["grasp the pink mug", "move the grasped object to the middle part of the hanger", "release"],
  "correct_plan": ["grasp pink mug", "move grasped pink mug to right part of hanger", "release pink mug"],
  "reward": 0
}

Key fields:

• reward: overall sample reward (0 = failure, 1 = success)
• planning_reward / execution_reward: indicates which stage failed
• failure_mode: fine-grained failure category
• failure_reason: natural language explanation of the failure
• old_* fields (planning only): the original instruction/plan before perturbation
• correct_plan (planning only): ground-truth correct plan

InternVL Fine-tuning Files (internVL_dataset_*.jsonl)

These files are formatted for direct fine-tuning of InternVL-style models. Each line is a JSON object containing a multi-turn conversation.

{
  "id": 350,
  "image": ["path/to/img1.png", "path/to/img2.png"],
  "height_list": [256, 256],
  "width_list": [256, 256],
  "conversations": [
    {"from": "human", "value": "Multiview images at the start of the subtask: ..."},
    {"from": "gpt", "value": "<answer> True </answer> <category> success </category>"}
  ]
}

Vanilla variants produce direct answers: <answer> boolean </answer> <category> category </category>

Thinking variants include chain-of-thought reasoning: <think> reasoning </think> <answer> boolean </answer> <category> category </category>

Image Details

• Resolution: 256 x 256 pixels
• Execution samples: 6 images (3 viewpoints x 2 timesteps). Viewpoints are front (0), left (1), and right (2).
• Planning samples: 1 image (front viewpoint of the initial scene)
• Format: PNG

UR5-Fail Specific Details

The UR5-Fail dataset was collected on a real tabletop setup using a 6-DoF UR5 robot arm equipped with a gripper and three Intel RealSense D435 cameras. The robot runs the 3D-LOTUS++ policy autonomously on 34 manipulation tasks. Subtasks are manually labeled as success or failure for execution data. Planning failures are generated using the automated perturbation method described in the paper.

Usage

Load the metadata to iterate over samples:

import json
from PIL import Image

with open("metadata_execution.jsonl") as f:
    for line in f:
        sample = json.loads(line)
        images = [Image.open(img_path) for img_path in sample["images"]]
        label = sample["execution_reward"]  # 1 = success, 0 = failure
        category = sample["failure_mode"]
        # ...

Load InternVL-format data for fine-tuning:

import json

with open("internVL_dataset_execution_thinking.jsonl") as f:
    for line in f:
        sample = json.loads(line)
        image_paths = sample["image"]
        conversations = sample["conversations"]
        # ...

Citation

If you use this dataset, please cite:

@inproceedings{
pacaud2025guardian,
title={Guardian: Detecting Robotic Planning and Execution Errors with Vision-Language Models},
author={Paul Pacaud and Ricardo Garcia Pinel and Shizhe Chen and Cordelia Schmid},
booktitle={Workshop on Making Sense of Data in Robotics: Composition, Curation, and Interpretability at Scale at CoRL 2025},
year={2025},
url={https://openreview.net/forum?id=wps46mtC9B}
}

License

This dataset is released under the Apache 2.0 license.