TY  - RPRT
TI  - SynthDemo-RL: Breaking the Zero-Reward Barrier in VLA Adaptation with LLM-Guided Synthetic Demonstrations
AU  - Hiroaki Kingetsu
AU  - Hiroaki Kurihara
AU  - Kaoru Yokoo
AU  - Kenji Fukumizu
AU  - Manohar Kaul
PY  - 2026
UR  - https://arxiv.org/abs/2609.21650
ID  - 2609.21650
ER  -