Datasets:

Modalities:
Image
License:
tracebench / website /environments /BounceBall /description.json
TommasoBendinelli's picture
TraceBench synchronized paper release
ec08d06
Raw
History Blame Contribute Delete
7.49 kB
{
"candidate_parameters": [
"weight of the mass",
"coefficient of viscous damping",
"coefficient of restitution of the right wall",
"coefficient of restitution of the left wall",
"inclination angle of the rail"
],
"download_link": "https://huggingface.co/datasets/eth-siplab/tracebench/tree/main/questions/BounceBall",
"environment_id": "BounceBall",
"name": "BounceBall",
"observed_channels": [
{
"id": "Initial_Spacer_F_x",
"label": "mass position along the rail",
"unit": ""
},
{
"id": "Initial_Spacer_F_v",
"label": "mass velocity along the rail",
"unit": ""
},
{
"id": "Right_Hard_Stop_f",
"label": "force on the right wall",
"unit": ""
}
],
"prompt_combinations": [
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nby simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R. The rail may be tilted by a fixed inclination angle. When the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail. When the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nReturn the final ranked answer directly.",
"desc_level": "high",
"task_type": "direct",
"training_samples": "none"
},
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nby simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R. The rail may be tilted by a fixed inclination angle. When the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail. When the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nUse the provided labeled examples as calibration examples for the same benchmark condition.\n\nReturn the final ranked answer directly.",
"desc_level": "high",
"task_type": "direct",
"training_samples": ">0"
},
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nby simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R. The rail may be tilted by a fixed inclination angle. When the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail. When the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nReturn executable analysis code and the final ranked answer.",
"desc_level": "high",
"task_type": "code",
"training_samples": "none"
},
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nby simulating a damped mass moving along a one-dimensional rail between two rigid walls located at positions x_L and x_R. The rail may be tilted by a fixed inclination angle. When the mass is not in contact with either wall, its motion along the rail is governed by viscous drag, proportional to velocity and opposite the direction of motion, and, when the inclination angle is nonzero, by the component of gravity along the rail. When the mass hits a wall, the collision is modeled as an instantaneous impact: the direction of motion reverses and the post-impact speed is reduced according to that walls's restitution coefficient.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nUse the provided labeled examples as calibration examples for the same benchmark condition.\n\nReturn executable analysis code and the final ranked answer.",
"desc_level": "high",
"task_type": "code",
"training_samples": ">0"
},
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nReturn the final ranked answer directly.",
"desc_level": "none",
"task_type": "direct",
"training_samples": "none"
},
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nUse the provided labeled examples as calibration examples for the same benchmark condition.\n\nReturn the final ranked answer directly.",
"desc_level": "none",
"task_type": "direct",
"training_samples": ">0"
},
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nReturn executable analysis code and the final ranked answer.",
"desc_level": "none",
"task_type": "code",
"training_samples": "none"
},
{
"agent_instruction": "You are given multivariate time-series observations from BounceBall.\n\nIdentify which physical parameter changed, or answer that no intervention occurred.\n\nThe observed channels are mass position along the rail, mass velocity along the rail, force on the right wall.\n\nUse the provided labeled examples as calibration examples for the same benchmark condition.\n\nReturn executable analysis code and the final ranked answer.",
"desc_level": "none",
"task_type": "code",
"training_samples": ">0"
}
],
"sample_count": 5,
"short_one_line_description": "A bouncing ball trajectory with impacts, velocity changes, and contact impulses."
}