CorrectAD: A Self-Correcting Agentic System to Improve End-to-end Planning in Autonomous Driving

arXiv 25.11 2025 53.6 system, application

TLDR

A self-correcting agentic system using diffusion-based video generation to improve end-to-end planning in autonomous driving by correcting failure cases.

Reasoning

The paper presents a novel pipeline combining a product manager agent and a generative video model (DriveSora) to generate failure-case data, showing strong empirical results on real-world datasets. However, the reliance on existing generative models and the lack of detailed analysis of generated data fidelity are potential weaknesses.

Read-first score

Read-first score 53.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 41.

Recency 8%

86.7

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

Topical relevance 42%

58.6

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 25%

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

Reproducibility 25%

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

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 402.

Keyword Scores

world model

video world model

generative world model

world simulator

world dynamics prediction

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

PM-Agent: an agent simulating a product manager to formulate data requirements for collecting data similar to failure cases
DriveSora: a generative model that produces spatiotemporally consistent videos aligned with 3D annotations
CorrectAD: a self-correcting agentic system that is end-to-end model-agnostic and improves any planner

Methodology

CorrectAD uses a PM-Agent to formulate data requirements for collecting data similar to failure cases. Then, a generative model (DriveSora) simulates data collection and annotation by generating spatiotemporally consistent videos conditioned on 3D layouts. The pipeline is end-to-end model-agnostic and can be applied to any planner.

Key Results

On nuScenes and an in-house dataset, CorrectAD corrects 62.5% and 49.8% of failure cases, reducing collision rates by 39% and 27% respectively.

Resources

Paper