Devil is in Narrow Policy: Unleashing Exploration in Driving VLA Models

• URL: http://arxiv.org/abs/2603.06049v1
• Date: Fri, 06 Mar 2026 09:01:34 GMT
• Title: Devil is in Narrow Policy: Unleashing Exploration in Driving VLA Models
• Abstract summary: Imitation Learning (IL) tends to collapse exploration and limit the potential of subsequent Reinforcement Learning stages.<n>We propose Curious-VLA, a framework that alleviates the exploit-explore dilemma through a two-stage design.<n>In the RL stage, we present Adaptive Diversity-Aware Sampling (ADAS) that prioritizes high-diversity samples and introduce Spanning Driving Reward (SDR)<n>On the Navsim benchmark, Curious-VLA achieves SoTA results (PDMS 90.3,S 85.4) and a Best-of-N PDMS of 94.8, demonstrating its effectiveness in unlocking the exploratory potential of
• Score: 31.413394596106926
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We identify a fundamental Narrow Policy limitation undermining the performance of autonomous VLA models, where driving Imitation Learning (IL) tends to collapse exploration and limit the potential of subsequent Reinforcement Learning (RL) stages, which often saturate prematurely due to insufficient feedback diversity. Thereby, we propose Curious-VLA, a framework that alleviates the exploit-explore dilemma through a two-stage design. During IL, we introduce a Feasible Trajectory Expansion (FTE) strategy to generate multiple physically valid trajectories and a step-wise normalized trajectory representation to adapt this diverse data. In the RL stage, we present Adaptive Diversity-Aware Sampling (ADAS) that prioritizes high-diversity samples and introduce Spanning Driving Reward (SDR) with a focal style weighting to amplify reward's value span for improving sensitivity to driving quality. On the Navsim benchmark, Curious-VLA achieves SoTA results (PDMS 90.3, EPDMS 85.4) and a Best-of-N PDMS of 94.8, demonstrating its effectiveness in unlocking the exploratory potential of VLA models. Code: https://github.com/Mashiroln/curious_vla.git.

Related papers

• CLEAR: Closed-Loop Reinforcement Learning at Scale for End-to-End Autonomous Driving [66.77752555295355]
  We present CLEAR, a system that enables closed-loop training using Reinforcement Learning (RL) at scale for E2E-AD.<n>We show that with a simple reward, CLEAR significantly outperforms previous methods and sets new state-of-the-art performance.
  arXiv  Detail & Related papers  (2026-07-03T00:34:50Z)
• Unleashing VLA Potentials in Autonomous Driving via Explicit Learning from Failures [14.313346858887286]
  Vision-Language-Action (VLA) models for autonomous driving often hit a performance plateau during Reinforcement Learning (RL) optimization.<n>This stagnation arises from exploration capabilities constrained by previous Supervised Fine-Tuning (SFT)<n>We propose Explicit Learning from Failures (ELF-VLA), a framework that augments RL with structured diagnostic feedback.
  arXiv  Detail & Related papers  (2026-03-01T11:41:22Z)
• Controllable Exploration in Hybrid-Policy RLVR for Multi-Modal Reasoning [88.42566960813438]
  CalibRL is a hybrid-policy RLVR framework that supports controllable exploration with expert guidance.<n>CalibRL increases policy entropy in a guided manner and clarifies the target distribution.<n>Experiments across eight benchmarks, including both in-domain and out-of-domain settings, demonstrate consistent improvements.
  arXiv  Detail & Related papers  (2026-02-22T07:23:36Z)
• DiffusionDriveV2: Reinforcement Learning-Constrained Truncated Diffusion Modeling in End-to-End Autonomous Driving [65.7087560656003]
  Generative diffusion models for end-to-end autonomous driving often suffer from mode collapse.<n>We propose DiffusionDriveV2, which leverages reinforcement learning to constrain low-quality modes and explore for superior trajectories.<n>This significantly enhances the overall output quality while preserving the inherent multimodality of its core Gaussian Mixture Model.
  arXiv  Detail & Related papers  (2025-12-08T17:29:52Z)
• Steering Vision-Language-Action Models as Anti-Exploration: A Test-Time Scaling Approach [78.4812458793128]
  We propose textbfTACO, a test-time-scaling framework that applies a lightweight pseudo-count estimator as a high-fidelity verifier of action chunks.<n>Our method resembles the classical anti-exploration principle in offline reinforcement learning (RL), and being gradient-free, it incurs significant computational benefits.
  arXiv  Detail & Related papers  (2025-12-02T14:42:54Z)
• Dual-Stream Diffusion for World-Model Augmented Vision-Language-Action Model [62.889356203346985]
  We propose DUal-STream diffusion (DUST), a world-model augmented VLA framework that handles the modality conflict.<n>DUST achieves up to 6% gains over a standard VLA baseline and implicit world-modeling methods.<n>On real-world tasks with the Franka Research 3, DUST outperforms baselines in success rate by 13%.
  arXiv  Detail & Related papers  (2025-10-31T16:32:12Z)
• ActiveVLN: Towards Active Exploration via Multi-Turn RL in Vision-and-Language Navigation [57.399685080574756]
  Existing MLLM-based VLN methods rely on imitation learning (IL) and often use DAgger for post-training.<n>We propose ActiveVLN, a VLN framework that explicitly enables active exploration through multi-turn RL.<n>Experiments show that ActiveVLN achieves the largest performance gains over IL baselines compared to both DAgger-based and prior RL-based post-training methods.
  arXiv  Detail & Related papers  (2025-09-16T03:31:46Z)
• IRL-VLA: Training an Vision-Language-Action Policy via Reward World Model [19.141499640543138]
  IRL-VLA is a novel close-loop Reinforcement Learning via textbfInverse textbfReinforcement textbfLearning reward world model with a self-built VLA approach.<n>In this paper, we introduce IRL-VLA, a novel close-loop Reinforcement Learning via textbfInverse textbfReinforcement textbfLearning reward world model with a self-built VLA approach.
  arXiv  Detail & Related papers  (2025-08-07T06:30:05Z)
• Divide-and-Conquer for Lane-Aware Diverse Trajectory Prediction [71.97877759413272]
  Trajectory prediction is a safety-critical tool for autonomous vehicles to plan and execute actions. Recent methods have achieved strong performances using Multi-Choice Learning objectives like winner-takes-all (WTA) or best-of-many. Our work addresses two key challenges in trajectory prediction, learning outputs, and better predictions by imposing constraints using driving knowledge.
  arXiv  Detail & Related papers  (2021-04-16T17:58:56Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.