ActPrune: Action-Grounded Visual Token Pruning for Efficient Vision-Language-Action Models
Overview
This research presents ActPrune, a training-free method that grounds visual token pruning in geometric priors derived from robot actions to reduce inference costs in vision-language-action (VLA) models. By leveraging action chunks predicted at the previous control step, the method retains relevant patches in fixed third-person views while pruning unchanged regions in wrist views after compensating for camera motion. Experiments on simulation environments (LIBERO and VLABench) and four real-robot tasks demonstrate up to 1.61x speedup and 59.5% FLOPs reduction with negligible degradation in success rates.
Generated from attributed reports · Updated 20 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 · speedup: 1.61 other · Basis not reported
“trate that \ours achieves up to 1.61 $\times$ speedup and 59.5\% FLOPs reduction with negligible degradation in success rates.”
Exact source · revision 1Source owner not reported
Report timeline
Follow attributed reports and material updates.
- arXiv Robotics — research abstractsSignalActPrune: Action-Grounded Visual Token Pruning for Efficient Vision-Language-Action Models
This research presents ActPrune, a training-free method that grounds visual token pruning in geometric priors derived from robot actions to reduce inference costs in vision-language-action (VLA) models. By leveraging action chunks predicted at the previous control step, the method retains relevant patches in fixed third-person views while pruning unchanged regions in wrist views after compensating for camera motion. Experiments on simulation environments (LIBERO and VLABench) and four real-robot tasks demonstrate up to 1.61x speedup and 59.5% FLOPs reduction with negligible degradation in success rates.
Event coverage history
Current attention 6·Peak within the comparable range 10（2026-11-09 06:00 UTC）·Change within the comparable range over 24 hours –
The trend compares participants observed continuously; its scope may be narrower than current attention totals. Hover or click to explore hourly observations, or use the left and right arrow keys.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).