Study Examines Sample Efficiency of Finetuning Vision-Language-Action Models
A new arXiv paper investigates how many demonstrations are needed when finetuning pretrained vision-language-action (VLA) models, which show strong zero-shot performance but are usually post-trained on demonstration data before deployment.
Read the full story at arXiv cs.CV ↗
Timeline · 1 report
- 2026-11-10 05:00 · arXiv cs.CV
How Many Demonstrations Does a VLA Need? On the Sample Efficiency of VLA Finetuning