DiVeR: Decision-Critical Verifier Learning for VLA Test-Time Scaling
Authors: Seongheon Park, Heecheol Kim, Shulin Tian, Lilika Makabe, Namiko Saito, Katsushi Ikeuchi, Sharon Li, Yasuyuki Matsushita
Organizations: University of Wisconsin–Madison · Microsoft Research Asia–Tokyo · Nanyang Technological University · The University of Osaka · The University of Tokyo
Scaling robot data and model capacity has improved Vision-Language-Action (VLA) policies, but further progress is constrained by the high cost of robotic data. Verifier-guided test-time scaling offers an efficient alternative by sampling multiple action candidates and selecting the one most likely to lead to task success at inference time. Existing classification-based verifiers learn from trajectory-level outcomes but treat all visited states equally, even though their value for candidate discrimination can vary across a trajectory. At many states, plausible actions are similar and provide limited discrimination signal, while only a sparse set of decision-critical states admits meaningfully different actions that can substantially affect downstream outcomes. To address this, we propose DiVeR, which estimates decision criticality from the dispersion of sampled action representations. DiVeR then uses this signal to reweight verifier learning toward states where action selection is most consequential, without requiring step-level annotations or additional environment interaction. Across LIBERO, RoboCasa, and real-world experiments on a Franka Research 3 robot, DiVeR consistently improves task success through more effective verifier-guided action selection, while adding negligible verifier inference overhead.
Figures & tables
Figure 1: Not all states matter equally for verifier learning. (a) At routine states , sampled action candidates remain similar, so distinguishing among them is less important for downstream behavior. At decision-critical states , candidates diverge more strongly, making accurate selection more consequential for task outcomes. (b) Across 500 LIBERO-Long trajectories from a frozen π0 policy, states with high decision criticality , measured by ut ( Equation 8 ), are sparse and temporally localized. These observations motivate focusing verifier learning on sparse decision-critical states where action selection matters most.
Figure 2: Overview of DiVeR . (a) At each visited state, we sample K action candidates from the frozen VLA policy and estimate decision criticality ut from the dispersion of their action-expert representations. The scores are standardized within each trajectory and converted into decision weights wt , which are paired with the executed action representation zt and trajectory outcome yt to construct the training dataset D . (b) The verifier is trained with decision-weighted binary cross-entropy, assigning greater importance to states with larger wt . (c) At inference, the verifier scores N sampled action candidates at each decision step and executes the highest-scoring action chunk.
Figure 3: Sampled action representations. Candidate representations are concentrated at a routine state (e.g., free-space motion) but more dispersed at a decision-critical state (e.g., before grasping).
Method
Policy: π0
Policy: π0.5
N=4
N=8
N=16
N=32
N=4
N=8
N=16
N=32
Baseline
83.5
93.5
KDPE ( Rosasco et al., 2025 )
83.5
84.5
82.8
82.3
94.0
92.7
94.5
95.7
MG-Select ( Jang et al., 2026 )
82.0
80.8
83.3
86.3
94.2
95.0
96.5
96.8
TACO ( Yang et al., 2025 )
84.2
83.0
83.5
85.0
94.5
94.8
95.5
95.7
SVM ( Tsao et al., 2026 )
82.8
85.2
85.5
86.5
94.0
95.5
94.0
97.0
Table 1: Average success rates (%) on LIBERO-Long with varying numbers of action candidates N sampled at test time. Bold and underlined values denote the best and second-best results, respectively.
Method
Policy: π0
Policy: π0.5
Articulated
Control
Object-Centric
Navigation
Average
Articulated
Control
Object-Centric
Navigation
Average
Baseline
37.2
16.2
42.5
0.0
32.2
50.0
18.8
58.3
0.0
43.1
KDPE
32.8
26.2
49.2
0.0
35.0
47.8
20.0
57.5
0.0
43.3
MG-Select
29.3
22.5
36.7
0.0
28.6
53.6
33.8
66.7
15.0
51.3
TACO
37.1
28.8
35.8
0.0
32.7
52.1
26.2
53.3
0.0
43.9
SVM
37.2
26.2
40.8
0.0
33.9
55.0
20.0
60.8
15.0
46.9
Table 2: Category-wise average success rates (%) on RoboCasa atomic tasks. Bold and underlined values denote the best and second-best results, respectively.
Table 3: Real-robot evaluation with π0.5 . Left : robot setup. Center : the four evaluation tasks. Right : task success rates (%) over 24 evaluation episodes per task.
Table 7
Figure 4: Cross-task generalization. Success rates (%) when the verifier is trained on one task category and evaluated on another.
Figure 5: Qualitative comparison of verifier-guided action selection. Starting from the same decision-critical state, we compare trajectories produced by a uniformly trained verifier and DiVeR (Ours) . In both (a) drawer closing and (b) block stacking, the two verifiers select different action candidates from the same state, leading to different downstream outcomes. Uniform verifier training selects actions that result in task failure, whereas DiVeR selects more appropriate candidates and successfully completes the task. Red circles highlight the resulting behaviors after these action selections.
Figure 6: Verifier architecture.
Appendix figures & tables11 assets
Supplementary material from the paper’s appendix.
Appendix
Hyperparameter
Search Space
Optimization
Optimizer
{Adam, AdamW , SGD}
Learning rate
{1e-2, 5e-3, 1e-3, 5e-4 , 1e-4}
Training epochs
{10, 30 , 50, 100}
Architecture
MLP hidden size
{ 64 , 256, 512}
Appendix
Table 7: Hyperparameter search space applied uniformly across all experiments. Selected values are bolded .
Table 8: Dataset statistics. Number of successful and failed episodes and their corresponding environment steps in the offline rollout datasets collected with (a) π0 and (b) π0.5 .
ID
Task
Skill
1
Place the green block in the bowl
Pick & Place
2
Place the cup on the plate
Pick & Place
3
Stack the red block on the blue block
Stacking
4
Close the drawer
Articulated Manipulation
Appendix
Table 9: Real-world robot evaluation tasks. We evaluate on four manipulation tasks spanning object placement, stacking, and articulated-object interaction.
ID
Task
1
Put both the alphabet soup and the tomato sauce in the basket
2
Put both the cream cheese box and the butter in the basket
3
Turn on the stove and put the moka pot on it
4
Put the black bowl in the bottom drawer of the cabinet and close it
5
Put the white mug on the left plate and put the yellow and white mug on the right plate
6
Pick up the book and place it in the back compartment of the caddy
Appendix
Table 10: Tasks in the LIBERO-Long benchmark. We evaluate on all 10 long-horizon manipulation tasks in the standard LIBERO-Long suite.
Category
Task
RoboCasa Skill
Articulated
Close blender lid
Open/Close Lid
Close fridge
Open/Close Door
Close toaster oven door
Open/Close Door
Open cabinet
Open/Close Door
Open drawer
Open/Close Drawer
Open stand mixer head
Open/Close Lid
Appendix
Table 11: RoboCasa evaluation tasks. The 18 Atomic-Seen tasks span four broad interaction categories: Articulated (7), Control (4), Object-Centric (6), and Navigation (1). These categories consolidate the finer-grained atomic skills defined by RoboCasa365.
Figure 7: Measured criticality ut vs. oracle criticality Coracle .
Method
Success (%)
Base
52.5
MGS
54.5
SVM
55.0
Ours
55.5
Appendix
Table 12: Success rate (%) with OpenVLA using N=4 candidates on LIBERO-Long.
Figure 8: Chunk-level action sampling latency.
K
Success (%)
4
52.8
16
53.9
32
54.7
64
55.0
Appendix
Table 13: Effect of the number of samples K used for criticality estimation.
Table 14: Representation extraction ablation on RoboCasa with π0.5 . Success rate (%) under different (a) action-token positions and (b) action-expert layers.
Figure 9: t-SNE visualization of action-expert representations. We visualize action-expert representations from a frozen π0 policy over 500 LIBERO-Long episodes. (a) Different colors denote different tasks, revealing task-dependent structure in the representation space. (b) Successful and failed trajectories occupy distinct regions of the representation space.
Vision-Language-Action models (VLAs) have demonstrated remarkable performance in robot control. However, they remain fundamentally limited in tasks that require high precision due to their single-inference paradigm. While test-time scaling approaches using external verifiers have shown promise, they require additional training and fail to generalize to unseen conditions. We propose Masking Distribution Guided Selection (MG-Select), a novel test-time scaling framework for VLAs that leverages the model's internal properties without requiring additional training or external modules. Our approach utilizes KL divergence from a reference action token distribution as a confidence metric for selecting the optimal action from multiple candidates. We introduce a reference distribution generated by the same VLA but with randomly masked states and language conditions as inputs, providing action uncertainty while remaining aligned with the target task distribution. Additionally, we propose a joint training strategy that enables the model to learn both conditional and unconditional distributions by applying dropout to state and language conditions, thereby further improving the quality of the reference distribution. Our experiments demonstrate that MG-Select provides a reliable reference for action selection through task-relevant condition masking and consistently improves base models across diverse simulation and real-world benchmarks.
Test-time sampling promises to turn one action prediction into a set of distinct executable candidates that a verifier can choose among. Downstream fine-tuning works against that promise. A vision-language-action model trained to solve one task distribution concentrates its flow-matching decoder around the dominant demonstrated action, so repeated samples differ only marginally and the verifier receives many copies of one decision. We call this proposal contraction and identify it as a key constraint on test-time scaling: repeated samples can remain highly redundant as the candidate budget grows. Action Dynamic Sampling (ADS) alleviates the constraint on the proposal side, during training. At each training step, ADS integrates K additional trajectories from the current policy under fresh base noise, scores each by the flow support along its path with a Hutchinson trace estimator, and corrects the policy's local action distribution toward this weighted set through a detached finite-set KL term. Training cost rises; deployment cost is unchanged. Across simulation and four physical-robot tasks, ADS raises proposal spread without the success penalty that Gaussian perturbation incurs, lifts oracle best-of-8 success on RoboCasa from 57.6% to 63.0%, improves fixed verifiers by 2.4 to 4.9 points in simulation, and raises TokenPB by up to 13.0 points on real-world manipulation.
Chen Zhao, Zipeng Ye, Haitao Shen +3
Renmin University of China, Beijing 100872, China · Key Laboratory of Data Engineering and KnowledgeEngineering, MOE, China · Meituan, Beijing 100102, China
Vision-language-action (VLA) models have shown strong promise for robotic manipulation, but their reliability at test time remains limited by one-shot action prediction, where even small action errors can cause grasp failure, collision, or incorrect task progression. A natural alternative is to equip VLA systems with test-time verification, allowing multiple candidate actions to be proposed and evaluated before execution. However, reliable action verification is challenging because it requires not only distinguishing subtle geometric differences between candidate actions, but also assessing whether an action makes meaningful progress toward the task goal. We present VeriSpace, a 3D-aware action verifier for test-time action selection in VLA systems. VeriSpace evaluates candidate actions through two key components: Dual-Path 3D-Injected Scene Encoding, which constructs a scene representation that jointly preserves visual semantics and explicit 3D geometry, and Spatially-Grounded Action Reasoning, which evaluates each action by reasoning over task-relevant spatial relations, geometric validity, and expected goal progress. Together, these components enable more reliable discrimination between subtle yet outcome-critical action candidates while remaining fully compatible with existing VLA policies. Experiments on public benchmarks and real-world robotic manipulation tasks show that VeriSpace consistently improves decision reliability over both underlying VLA policies and prior verification-based methods, yielding substantial gains in both in-distribution and out-of-distribution settings.
Guiyu Zhao, Longteng Guo, Junyou Zhu +6
Institute of Automation, Chinese Academy of Sciences, Beijing, China · University of Chinese Academy of Sciences, Beijing, China