Event
DiVeR: Reweighting VLA Verifiers with Decision Criticality
1 reports · 1 sources
Seongheon Park and co-authors propose the DiVeR verifier learning method: it estimates "decision criticality" from the dispersion of sampled action representations, and uses this to reweight verifier learning, focusing learning on the states where action selection has the greatest impact. The authors claim that this method requires no step-level annotation or additional environment interaction. In the LIBERO and RoboCasa benchmarks and real-robot experiments on the Franka Research 3, DiVeR improves task success rates through verifier-guided action selection, and the authors claim that the verifier's inference overhead is negligible.
Report timeline
- DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling
DiVeR estimates decision criticality from the dispersion of sampled action representations and uses it to reweight verifier learning, focusing learning on states where action selection matters most, without requiring step-level annotations or additional environment interaction. In experiments on LIBERO, RoboCasa, and a Franka Research 3 real robot, DiVeR consistently improves task success rates by guiding action selection with a more effective verifier, and verifier inference overhead is negligible.
arXiv cs.RO ·