MaineCoon: Pursuing A Real-Time Audio-Visual Social World Model

arXiv 2026 54.4 method, system

TLDR

MaineCoon is a 22B parameter real-time audio-visual autoregressive model for social world modeling, achieving 47.5 FPS on a single GPU.

Reasoning

The paper introduces a novel social world model with innovative training techniques and real-time interaction capabilities, addressing a gap in human-centric social dynamics. However, the abstract lacks explicit details on real-world evaluation or benchmarks, and the concept of social world models is not yet validated against existing standards.

Read-first score

Read-first score 54.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 42.

Recency 6%

100

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

Citation impact 18%

95.2

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.95220779

Topical relevance 29%

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

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: 484.

Keyword Scores

world model

interactive world model

generative world model

video world model

world simulator

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

First real-time audio-visual autoregressive model with 22B parameters for social-interactive applications
Self-resampling technique for efficient training
Cross-modal representation alignment
Domain-aware preference optimization
Reinforced online-policy distillation (ROPD)
Agentic streaming inference framework with agentic cache management and prompt planning for thousand-second-scale generation

Methodology

MaineCoon is a 22B-parameter real-time audio-visual autoregressive model designed for streaming generation and sub-second interaction. Training employs novel techniques including self-resampling, cross-modal representation alignment, domain-aware preference optimization, and reinforced online-policy distillation. Inference uses an agentic streaming framework with cache management and prompt planning to mitigate drift over long horizons.

Key Results

Achieves a record-breaking frame rate of up to 47.5 FPS on a single GPU, setting a new state-of-the-art benchmark for high-quality, low-latency, and long-horizon audio-visual autoregressive models.

Resources

Paper