UniviewVLA: A Unified Multiview Vision-Language-Action Model with World Modeling

arXiv 2026 59.4 method, application

TLDR

UniviewVLA uses a world model to generate multiview future views from two cameras, improving robot manipulation in occluded tasks without extra hardware.

Reasoning

The paper presents a novel approach to handle occlusion by generating future multiview views, with efficient token compression and view selection. However, it relies on standard two-camera input and benchmarks that may not fully capture real-world occlusion complexity.

Read-first score

Read-first score 59.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 50.

Recency 6%

100

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

Citation impact 18%

94.4

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.94368595

Topical relevance 29%

71.4

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

Methodology quality 18%

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

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: 474.

Keyword Scores

world model

generative world model

world dynamics prediction

video world model

interactive world model

model-based reinforcement learning world model

world simulator

Deep Analysis

Innovations

Unified multiview Vision-Language-Action model with world modeling that infers multiview scene evolution from only standard two-camera observations
Motion-Informative Token Compression to reduce per-view latency from 6-7s to 0.2-0.3s by compressing each generated view from 625 to 16 tokens
Training-free Action-Entropy View Selection that dynamically identifies the most action-informative view at different inference stages

Methodology

UniviewVLA uses a world model to generate multiview future views from standard two-camera observations (agent-view and wrist-view), then applies Motion-Informative Token Compression to reduce token count per view and Action-Entropy View Selection to choose the most informative view for action prediction. The model is trained and evaluated on standard occlusion-free benchmarks (LIBERO, CALVIN) and customized occlusion-focused tasks, with real-robot experiments.

Key Results

UniviewVLA achieves 95.8% on LIBERO and 4.60 on CALVIN ABCD to D on standard occlusion-free benchmarks; on customized occlusion-focused tasks, it improves success rate from 40.0% to 73.3%, and average real-robot success rate by 33.4 points.

Resources

Paper