Network-Efficient World Model Token Streaming

arXiv 2026 48.7 method, application

TLDR

Proposes an adaptive keyframe-delta streaming protocol for discrete world model tokens, improving rate-distortion and perplexity under packet loss.

Reasoning

Strengths include a novel online algorithm for efficient streaming of world model states with clear experimental improvements. Weaknesses are the narrow driving domain focus and lack of comparison to other compression methods or full world model evaluation.

Read-first score

Read-first score 48.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 31.

Recency 6%

100

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

Methodology quality 18%

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

Citation impact 18%

60.8

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.60793858

Topical relevance 29%

44.3

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=code

Citation velocity 12%

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

Field roles

FrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 230.

Keyword Scores

world model

generative world model

world dynamics prediction

video world model

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Stride-16 VQ-U-Net tokenizer with codebook size 8,192 mapping 288x512 frames to 18x32 token grids (576 tokens/frame, 936 bytes/frame under fixed-length coding)
Fully online, label-free keyframe-delta protocol that prioritizes delta updates via cosine distance in codebook embedding space and triggers keyframes adaptively using a Hamming-drift threshold
Consistent improvement in rate distortion frontier over periodic keyframes at matched bitrates, with up to 7.2% reduction in embedding distortion and 6.3% improvement in perplexity

Methodology

The paper proposes a stride-16 VQ-U-Net tokenizer that maps each 288x512 frame to an 18x32 grid of token IDs (576 tokens/frame) using a codebook of size 8,192, resulting in 936 bytes/frame under fixed-length coding. A keyframe-delta protocol is employed under strict per-message payload budgets and packet loss, with a fully online, label-free algorithm that prioritizes delta updates via cosine distance in codebook embedding space and triggers keyframes adaptively using a Hamming-drift threshold. The method is evaluated against periodic keyframes at matched bitrates using rate distortion metrics and perplexity of a lightweight next-token predictor conditioned on receiver states.

Key Results

At 0.024 Mb/s (200-byte budget), dynamic-only embedding distortion drops from 0.0712 to 0.0661 (7.2% improvement); at 0.036 Mb/s (400-byte budget) from 0.0427 to 0.0407 (4.8%). Under 10% delta packet loss at 200 bytes, dynamic-only distortion is 0.0757 versus 0.0789 for periodic baseline. Perplexity improves from 206.0 to 193.1 (6.3%) at 0.024 Mb/s and from 158.9 to 155.6 (2.1%) at 0.036 Mb/s.

Resources

Paper