Pre-VLA: Preemptive Runtime Verification for Reliable Vision-Language-Action and World-Model Rollouts

arXiv 2026 52.1 method

TLDR

Pre-VLA preemptively verifies action validity for VLA and world models, improving success rates and reducing errors on LIBERO.

Reasoning

The paper presents a novel runtime verification method with strong empirical results on a standard benchmark, addressing a practical deployment issue. However, the abstract lacks details on generalization to diverse tasks and environments, and the method's reliance on a specific backbone may limit applicability.

Read-first score

Read-first score 52.1, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 37.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Citation impact 18%

83.6

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.83577566

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,experiment

Topical relevance 29%

52.9

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridge

Rank sensitivity

Stability: volatile; rank range: 322.

Keyword Scores

world model

generative world model

world dynamics prediction

world simulator

model-based reinforcement learning world model

interactive world model

video world model

Deep Analysis

Innovations

Preemptive runtime verification before physical execution or world-model imagination
Efficient multimodal backbone with modality-aware pooling and lightweight dual-branch head for safety confidence and advantage score prediction
Multi-task training objective combining Focal classification, advantage regression, and soft-threshold calibration to handle class imbalance and boundary decisions
Dual-mode preemptive resampling scheduler that filters low-quality actions and triggers adaptive resampling under limited computation budget

Methodology

Pre-VLA uses an efficient multimodal backbone with modality-aware pooling and a lightweight dual-branch head to predict safety confidence and critic-derived advantage scores for candidate action chunks. It is trained with a multi-task objective combining Focal classification, advantage regression, and soft-threshold calibration. During deployment, a dual-mode preemptive resampling scheduler filters low-quality actions and triggers adaptive resampling under a limited computation budget.

Key Results

On the LIBERO benchmark, Pre-VLA improves the average closed-loop success rate from 30.79% to 37.62% over RynnVLA-002, reduces task execution steps, achieves 183.9 ms average forward verification time per action chunk, and mitigates error accumulation in world-model rollouts.

Resources

Paper