We'll Fix it in Post: Improving Text-to-Video Generation with Neuro-Symbolic Feedback

arXiv 2025 22.2 method, system

TLDR

Introduces NeuS-E, a zero-training video refinement pipeline using neuro-symbolic feedback to improve text-to-video alignment for complex prompts, boosting temporal/logical consistency by ~40%.

Reasoning

The paper addresses a practical limitation in text-to-video generation with a novel zero-training refinement approach and demonstrates strong empirical gains across multiple models. However, the abstract lacks details about the neuro-symbolic feedback mechanism and evaluation methodology, making it hard to assess generalizability and robustness.

Read-first score

Read-first score 22.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 0.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%

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

Reproducibility 25%

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

Topical relevance 42%

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

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 223.

Keyword Scores

world model

world simulator

generative world model

interactive world model

video world model

world dynamics prediction

model-based reinforcement learning world model

Resources

Paper