Driving · 2026-10-20

EvoAD: Self-Evolving Autonomous Driving via Latent World-Model Guided Post-hoc Reflection

Junjie Yang, Qiming Zhang, Yuyan Li, Shuqi Shen, Yusong Huang, Yanggang Sheng, Wenpeng Xu, Ziyi Shi, Tianshun Li, Yijie Chen, Yujia Wang, Jun Ma, Xinhu Zheng, Hongliang Lu, Hai Yang

arXiv:2610.23886 PDF

AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.

TL;DR

EvoAD is a self-evolving end-to-end autonomous driving framework that uses latent world-model guided post-hoc reflection. It stores reflected driving experiences in a BEV feature memory pool, retrieves scene-relevant precedents by geometric/semantic similarity, and uses a latent world model to evaluate candidate trajectories in a compact planning-oriented space. The system refines future behavior via iterative reflection without retraining the base driving policy.

Why it matters

The authors address the common issue that end-to-end driving models are deployed as frozen policies, potentially repeating decisions and requiring costly data collection/retraining to fix failures. EvoAD aims to enable post-deployment improvement by turning executed outcomes into reusable improvement signals, using latent world-model evaluation plus retrieval-guided refinement.

Method

Stores reflected driving experiences in a BEV feature memory pool and retrieves scene-relevant precedents using geometric and semantic similarity.
Uses a latent world model to extract future latents conditioned on candidate trajectories, which are used for future-aware trajectory evaluation and reranking.
From reflection round 2, uses a retrieval-guided planning-aware refiner to produce feedback-conditioned residual corrections based on prior post-hoc feedback, iterating across reflection rounds.

Limitation

The paper states: “We have validated it on NAVSIM v1, v2, and the reactive NavHard split. In future work, we will extend it to more benchmarks and simulators such as CARLA (Dosovitskiy et al., 2017). Stronger latent world models and uncertainty-aware retrieval may further improve refinement reliability.”

Abstract (from arXiv)

End-to-end autonomous driving systems have achieved rapid progress by learning unified policies that map sensor observations directly to driving trajectories. However, once deployed, these systems typically remain frozen: they may repeat similar decisions in similar scenes, and failures can only be addressed through costly data collection and retraining. This sharply contrasts with human drivers, who improve lifelong through experience, memory, and reflection. In this work, we ask whether autonomous driving systems can achieve similar post-deployment self-evolution to human drivers. We propose EvoAD, a self-evolving autonomous driving framework via latent world-model guided post-hoc reflection. EvoAD stores driving experiences in a bird's-eye-view feature memory pool, retrieves scene-relevant precedents according to geometric and semantic similarity, and uses a latent world model to evaluate the future consequences of candidate trajectories in a compact planning-oriented space. Instead of treating memory as external textual contexts or using world models only for pre-action simulation, EvoAD transforms executed driving outcomes into reusable improvement signals for future decisions. This enables the system to diagnose how prior trajectories could have been improved and to refine behavior in similar future scenarios without retraining the base driving policy. Experiments on NAVSIM v1 and v2 show consistent improvements across reflection rounds and state-of-the-art performance, demonstrating the potential of world models as reflective mechanisms for post-deployment self-evolution in autonomous driving.

Publication record

Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.

First public release: 2026-09-26 · Indexed by Pub-AI: 2026-10-20 UTC

preprint