Thesis Supplementary Materials
Vision-Language-Action (VLA) Results
← Back to Home
01 —
Untrained VLA Models
Untrained SmolVLA
Completely untrained SmolVLA model. Note the erratic behavior.
Dipping Behavior
Initial evaluation run with model trained on 25 samples demonstrating dipping behavior.
02 —
Trained VLA Models
Successful VLA Trial
End-to-end mission demonstration illustrating successful execution of VLA grasp. Trained on all 127 samples.
03 —
Training Dataset
new_libero · HuggingFace
LIBERO pick-and-place demonstrations used to fine-tune the VLA model.
Hosted on HuggingFace Datasets — Xavier033/new_libero.