VideoCoCo: Code-as-CoT for Physically-Consistent Video Generation via an Agentic Dual-Engine System

arXiv 2026 42.8 method, system

TLDR

VideoCoCo uses executable Blender code as a chain-of-thought to generate physically consistent videos via a dual-engine framework.

Reasoning

The paper introduces a novel approach separating process-level reasoning (code simulation) from visual realization, achieving strong benchmark results. However, it does not discuss limitations or generalizability, and the reliance on a specific simulator may limit applicability.

Read-first score

Read-first score 42.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 35.

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,benchmark,dataset,result

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

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code,dataset

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 486.

Keyword Scores

world simulator

video world model

world model

world dynamics prediction

generative world model

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Agentic dual-engine framework using executable Blender code as a process-level chain-of-thought for physically consistent video generation
Decomposition into a coding agent that synthesizes an executable Blender program and a generative video engine that transforms the deterministic spatiotemporal draft into photorealistic video via draft-conditioned editing
Curated dataset VideoCoCo-3K of draft-instruction-target triplets to adapt the video editor to simulated drafts

Methodology

Given a text prompt, a coding agent generates a Blender program that explicitly specifies scene geometry and temporal evolution, which is executed by a simulation engine to produce a deterministic spatiotemporal draft; a generative video engine then performs draft-conditioned editing to convert the draft into a photorealistic video. A dataset of draft-instruction-target triplets (VideoCoCo-3K) is constructed to adapt the editor to simulated drafts.

Key Results

VideoCoCo improves the OmniWeaving baseline from 0.475 to 0.558 on PhyGenBench and from 52.18 to 77.88 on VBench-2.0, achieving the best average score on both benchmarks.

Resources

Paper