Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,39 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
tags:
|
| 4 |
+
- clarus
|
| 5 |
+
- vljepa
|
| 6 |
+
- video-inference
|
| 7 |
+
- structure
|
| 8 |
+
- alignment
|
| 9 |
+
- nonpixel
|
| 10 |
+
---
|
| 11 |
+
|
| 12 |
+
VL-JEPA Physics Consistency Probe (v01)
|
| 13 |
+
|
| 14 |
+
Tests if a model can track basic physical expectations from partial video context.
|
| 15 |
+
|
| 16 |
+
This dataset checks:
|
| 17 |
+
• continuity over frames
|
| 18 |
+
• cause and effect
|
| 19 |
+
• prevention of impossible events
|
| 20 |
+
• where inference must stop
|
| 21 |
+
|
| 22 |
+
Columns:
|
| 23 |
+
video_scenario
|
| 24 |
+
given_frames
|
| 25 |
+
physics_principle
|
| 26 |
+
expected_inference
|
| 27 |
+
failure_mode
|
| 28 |
+
stop_condition
|
| 29 |
+
|
| 30 |
+
Use cases:
|
| 31 |
+
• early signal of inference drift
|
| 32 |
+
• baseline for robotics pre-training
|
| 33 |
+
• measurement of hidden confabulation
|
| 34 |
+
|
| 35 |
+
Evaluation goal:
|
| 36 |
+
Can the model keep contact with physical reality without inventing states?
|
| 37 |
+
|
| 38 |
+
Files:
|
| 39 |
+
vljepa_physics_consistency_probe_v01.csv
|