{"slug": "personadrive-human-style-retrieval-augmented-vla-agents-for-closed-loop-driving", "title": "PersonaDrive: Human-Style Retrieval-Augmented VLA Agents for Closed-Loop Driving Simulation", "summary": "Researchers have developed PersonaDrive, a pipeline that conditions vision-language-action driving agents on retrieved demonstrations from a style-instructed human driving dataset to produce aggressive, neutral, or conservative non-ego traffic agents in closed-loop simulations. The system, which requires no per-style retraining, improves driving scores by up to 4.6% over existing models on the Bench2Drive benchmark while enabling style-diverse behavior. Average speed and acceleration increase by 18% and 25% from conservative to aggressive instructions, demonstrating the pipeline's ability to generate human-like driving styles for more realistic simulation environments.", "body_md": "arXiv:2606.12616v1 Announce Type: new\nAbstract: Closed-loop driving simulators typically populate their environments with non-ego traffic agents that behave largely the same way, produced either by rule-based traffic managers or by learned models trained toward a single behavioral mode. Recent work introduces style variation through post-hoc labels on observational data or LLM-inferred reward weights, but these signals act as proxies for what a style should reward rather than demonstrations of humans explicitly asked to drive in that style. We introduce PersonaDrive, a pipeline that conditions a vision-language-action (VLA) driving agent on retrieved demonstrations from a style-instructed human driving dataset, in which participants drive CARLA leaderboard routes under aggressive, neutral, and conservative instructions on a driver-in-the-loop rig. The pipeline has three stages: (i) offline triplet mining over per-style human driving data using a combined image-text similarity score; (ii) training a lightweight retrieval head that fuses frozen visual features with a small control encoder over per-style databases; and (iii) fine-tuning a single VLA backbone to treat retrieved context points as in-context behavioral demonstrations during waypoint prediction. At inference, the same backbone is conditioned on any style by swapping which per-style database the retrieval head queries, so selecting a style requires no per-style retraining while enabling human-style, style-diverse non-ego agents for closed-loop simulation. On Bench2Drive, PersonaDrive (no style) improves the driving score by 4.6% over SimLingo and 2.5% over HiP-AD, and under style conditioning attains the highest driving score in every style within a roughly 2% band (its weakest style surpassing the strongest baseline, DMW, by 5.4%), while average speed and acceleration rise by 18% and 25% from the conservative to the aggressive instruction.", "url": "https://wpnews.pro/news/personadrive-human-style-retrieval-augmented-vla-agents-for-closed-loop-driving", "canonical_source": "https://arxiv.org/abs/2606.12616", "published_at": "2026-06-12 04:00:00+00:00", "updated_at": "2026-06-12 04:51:37.852488+00:00", "lang": "en", "topics": ["autonomous-vehicles", "artificial-intelligence", "machine-learning", "computer-vision", "ai-agents"], "entities": ["PersonaDrive", "CARLA", "VLA"], "alternates": {"html": "https://wpnews.pro/news/personadrive-human-style-retrieval-augmented-vla-agents-for-closed-loop-driving", "markdown": "https://wpnews.pro/news/personadrive-human-style-retrieval-augmented-vla-agents-for-closed-loop-driving.md", "text": "https://wpnews.pro/news/personadrive-human-style-retrieval-augmented-vla-agents-for-closed-loop-driving.txt", "jsonld": "https://wpnews.pro/news/personadrive-human-style-retrieval-augmented-vla-agents-for-closed-loop-driving.jsonld"}}