← Back to the feed
Energy & materialsResearch findings

RAD · A Retrieval-Augmented Framework for Meta-Action Decision-Making in Autonomous Driving

Original published 2025.03.18 12:25 KSTAdded to Blackdeck 2026.11.10 12:05 KST

What has changed?

A retrieval-augmented decision-making framework is proposed to address spatial perception limitations and hallucination in vision-language models for autonomous

Key points

  1. Accurately understanding and deciding high-level meta-actions is essential for reliable and safe autonomous driving systems.
  2. Vision-language models often suffer from inadequate spatial perception and hallucination in complex driving scenarios.
  3. The proposed RAD framework enhances decision accuracy through a retrieval-augmented generation pipeline.

Connect it to your work

Improves meta-action decision accuracy on a curated NuScenes-based dataset.

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivRAD: Retrieval-Augmented Decision-Making of Meta-Actions with Vision-Language Models in Autonomous Driving