Datasets:
scene_id stringlengths 9 16 | task stringclasses 3
values | split stringclasses 1
value | difficulty stringclasses 4
values | prompt stringlengths 223 555 | answer stringlengths 4 16 | reasoning stringlengths 84 269 | assistant_text stringlengths 133 314 | n_frames int64 1 8 | frames_b64 listlengths 1 8 | ground_truth stringlengths 227 14.5k | config stringlengths 497 1.14k |
|---|---|---|---|---|---|---|---|---|---|---|---|
ttc_00000 | ttc | train | hard | "Two objects are moving in this 0.6-second video.\n\nObject A: purple sphere\nObject B: white sphere(...TRUNCATED) | 0.78 | "The video covers 0.56s of motion before collision. Based on the closing speed observed across the 7(...TRUNCATED) | "<reasoning>The video covers 0.56s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 7 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 0.779, \"contact_force\": 65.6447, \"contact_(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"hard\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.15(...TRUNCATED) |
ttc_00001 | ttc | train | medium | "Two objects are moving in this 1.0-second video.\n\nObject A: orange sphere\nObject B: green sphere(...TRUNCATED) | 1.53 | "The video covers 0.99s of motion before collision. Based on the closing speed observed across the 8(...TRUNCATED) | "<reasoning>The video covers 0.99s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 8 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 1.531, \"contact_force\": 72.7956, \"contact_(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"medium\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.(...TRUNCATED) |
ttc_00002 | ttc | train | medium | "You are watching a short video clip of two objects moving in a physics scene.\nThe clip shows 0.9 s(...TRUNCATED) | 1.29 | "The video covers 0.94s of motion before collision. Based on the closing speed observed across the 8(...TRUNCATED) | "<reasoning>The video covers 0.94s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 8 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 1.287, \"contact_force\": 311.4962, \"contact(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"medium\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.(...TRUNCATED) |
ttc_00003 | ttc | train | medium | "This video shows two objects over 0.5 seconds (6 frames at 10 fps).\n\nObject A: green box | Obje(...TRUNCATED) | 0.73 | "The video covers 0.48s of motion before collision. Based on the closing speed observed across the 6(...TRUNCATED) | "<reasoning>The video covers 0.48s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 6 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 0.735, \"contact_force\": 95.0344, \"contact_(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"medium\", \"obj1\": {\"shape\": \"box\", \"size\": 0.147(...TRUNCATED) |
ttc_00004 | ttc | train | medium | "Two objects are moving in this 0.6-second video.\n\nObject A: orange sphere\nObject B: purple cylin(...TRUNCATED) | 0.98 | "The video covers 0.64s of motion before collision. Based on the closing speed observed across the 7(...TRUNCATED) | "<reasoning>The video covers 0.64s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 7 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 0.983, \"contact_force\": 113.274, \"contact_(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"medium\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.(...TRUNCATED) |
ttc_00005 | ttc | train | medium | "Two objects are moving in this 0.8-second video.\n\nObject A: orange sphere\nObject B: yellow spher(...TRUNCATED) | 1.29 | "The video covers 0.82s of motion before collision. Based on the closing speed observed across the 8(...TRUNCATED) | "<reasoning>The video covers 0.82s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 8 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 1.288, \"contact_force\": 103.1004, \"contact(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"medium\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.(...TRUNCATED) |
ttc_00007 | ttc | train | hard | "Two objects are moving in this 0.7-second video.\n\nObject A: yellow sphere\nObject B: green sphere(...TRUNCATED) | 0.99 | "The video covers 0.74s of motion before collision. Based on the closing speed observed across the 8(...TRUNCATED) | "<reasoning>The video covers 0.74s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 8 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 0.988, \"contact_force\": 165.0492, \"contact(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"hard\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.09(...TRUNCATED) |
ttc_00008 | ttc | train | easy | "You are watching a short video clip of two objects moving in a physics scene.\nThe clip shows 1.8 s(...TRUNCATED) | 2.34 | "The video covers 1.84s of motion before collision. Based on the closing speed observed across the 8(...TRUNCATED) | "<reasoning>The video covers 1.84s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 8 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 2.344, \"contact_force\": 64.065, \"contact_p(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"easy\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.16(...TRUNCATED) |
ttc_00010 | ttc | train | medium | "This video shows two objects over 1.1 seconds (8 frames at 10 fps).\n\nObject A: red cylinder | O(...TRUNCATED) | 1.41 | "The video covers 1.10s of motion before collision. Based on the closing speed observed across the 8(...TRUNCATED) | "<reasoning>The video covers 1.10s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 8 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 1.408, \"contact_force\": 98.9187, \"contact_(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"medium\", \"obj1\": {\"shape\": \"cylinder\", \"size\": (...TRUNCATED) |
ttc_00011 | ttc | train | hard | "You are watching a short video clip of two objects moving in a physics scene.\nThe clip shows 0.6 s(...TRUNCATED) | 0.78 | "The video covers 0.57s of motion before collision. Based on the closing speed observed across the 7(...TRUNCATED) | "<reasoning>The video covers 0.57s of motion before collision. Based on the closing speed observed a(...TRUNCATED) | 7 | ["/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAUDBAQEAwUEBAQFBQUGBwwIBwcHBw8LCwkMEQ8SEhEPERETFhwXExQaFRERGCEYGh(...TRUNCATED) | "{\"collision_occurred\": true, \"time_to_collision\": 0.778, \"contact_force\": 204.6763, \"contact(...TRUNCATED) | "{\"task_type\": \"ttc\", \"difficulty\": \"hard\", \"obj1\": {\"shape\": \"sphere\", \"size\": 0.09(...TRUNCATED) |
End of preview. Expand in Data Studio
PhysSim-VLM Dataset
Paper: Synthetic Physics as Supervision: Learning Real-World Physical Reasoning in Vision-Language Models
Venue: AI4Physics Workshop @ ICML 2026
Authors: Swastik R, Natesha B V (IIIT Raichur)
Dataset Description
PhysSim-VLM is a fully synthetic physics-reasoning dataset for training and evaluating vision-language models (VLMs). It contains 15,000 multi-frame scenes (train: 12,023 / val: 1,477 / test: 1,500) generated from two physics simulators:
- MuJoCo — rigid-body dynamics: time-to-collision (TTC), pile stability, projectile trajectory
- PhiFlow — continuum fluid simulation: flow direction, viscosity comparison, fluid level
Each example consists of an 8-frame video rollout of geometric objects (coloured boxes, spheres, cylinders) interacting under physical laws, paired with a free-text question and an answer derived directly from simulator ground-truth state — no human annotation involved.
Intended Use
- Fine-tuning VLMs on physics-grounded visual reasoning
- Studying synthetic-to-real transfer for physical reasoning
- Probing what physics concepts can be taught via simulator supervision alone
Dataset Structure
| Split | Size |
|---|---|
| train | 12,023 |
| val | 1,477 |
| test | 1,500 |
Fields
| Field | Type | Description |
|---|---|---|
scene_id |
string | Unique scene identifier |
task |
string | Task family (e.g., ttc, stability, trajectory, fluid_direction, fluid_viscosity, fluid_level) |
frames_b64 |
list[string] | 1–8 video frames encoded as base64 PNG strings |
reasoning |
string | Free-text chain-of-thought answer derived from simulator state |
config |
dict | Scene configuration (object properties, simulator parameters) |
Data Generation
Scenes are generated using:
- MuJoCo 3.x for rigid-body physics (collision detection, gravity, friction)
- PhiFlow for fluid simulation (Navier-Stokes incompressible flow)
Generation scripts are available in the project code repository:
https://github.com/Swastikr/PhysSim-VLM
Citation
@inproceedings{swastik2026physsim,
title = {Synthetic Physics as Supervision: Learning Real-World Physical Reasoning in Vision-Language Models},
author = {Swastik R and Natesha B V},
booktitle = {AI4Physics Workshop at ICML 2026},
year = {2026},
url = {https://huggingface.co/datasets/Swastikr/PhysSim-VLM-Dataset}
}
License
- Downloads last month
- 106