Behavioral Steering of Vision-Language-Action Models in the Low-Parameter Regime
Overview
This research investigates adapting pretrained vision-language-action models (VLAs) to specific behavioral preferences, such as carrying objects higher, by freezing a 7B parameter OpenVLA-OFT policy and optimizing TinyLoRA adapters with only 8 to 384 trainable parameters. Using reinforcement learning methods like GRPO or evolutionary search (CMA-ES), the authors achieve significant improvements in task performance—for instance, raising carry height by 120 mm on trained tasks at 92% success—while retaining most of the model's pretrained capabilities. The study highlights that adapter capacity and optimizer choice critically influence the trade-off between style gains and capability retention, showing that even minimal parameter updates can effectively steer VLA behavior.
Generated from attributed reports · Updated 1 hours ago
Event evidence and corrections
0 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · carry height increase on trained tasks: 120 other · Basis not reported
“OpenVLA-OFT 7B policy and optimize TinyLoRA adapters with 8 to 384 trainable parameters using either policy-gradient RL (GRPO) or evolutionary search (CMA-ES). With just 256 trainable parameters, we raise carry height by 120 mm on trained tasks at 92% success and by 74 mm on a held-out task at 80% success. By rescaling the learned update, we can tune the trade-off between style gains and capability retention without”
Exact source · revision 1Source owner not reported
Reported quantity · carry height increase on held-out task: 74 other · Basis not reported
“OpenVLA-OFT 7B policy and optimize TinyLoRA adapters with 8 to 384 trainable parameters using either policy-gradient RL (GRPO) or evolutionary search (CMA-ES). With just 256 trainable parameters, we raise carry height by 120 mm on trained tasks at 92% success and by 74 mm on a held-out task at 80% success. By rescaling the learned update, we can tune the trade-off between style gains and capability retention without”
Exact source · revision 1Source owner not reported
Reported quantity · trainable parameters in TinyLoRA adapters: 256 other · Basis not reported
“OpenVLA-OFT 7B policy and optimize TinyLoRA adapters with 8 to 384 trainable parameters using either policy-gradient RL (GRPO) or evolutionary search (CMA-ES). With just 256 trainable parameters, we raise carry height by 120 mm on trained tasks at 92% success and by 74 mm on a held-out task at 80% success. By rescaling the learned update, we can tune the trade-off between style gains and capability retention without”
Exact source · revision 1Source owner not reported
Reported quantity · trainable parameters range: 8 other · Basis not reported
“OpenVLA-OFT 7B policy and optimize TinyLoRA adapters with 8 to 384 trainable parameters using either policy-gradient RL (GRPO) or evolutionary search (CMA-ES). With just 256 trainable parameters, we raise carry height by 120 mm on trained tasks at 92% success and by 74 mm on a held-out task at 80% success. By rescaling the learned update, we can tune the trade-off between style gains and capability retention without”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalBehavioral Steering of Vision-Language-Action Models in the Low-Parameter Regime
This research investigates adapting pretrained vision-language-action models (VLAs) to specific behavioral preferences, such as carrying objects higher, by freezing a 7B parameter OpenVLA-OFT policy and optimizing TinyLoRA adapters with only 8 to 384 trainable parameters. Using reinforcement learning methods like GRPO or evolutionary search (CMA-ES), the authors achieve significant improvements in task performance—for instance, raising carry height by 120 mm on trained tasks at 92% success—while retaining most of the model's pretrained capabilities. The study highlights that adapter capacity and optimizer choice critically influence the trade-off between style gains and capability retention, showing that even minimal parameter updates can effectively steer VLA behavior.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).