Towards Unified World Models for Visual Navigation via Memory-Augmented Planning and Foresight

arXiv:2510.08713 · cs.AI, cs.CV, cs.RO · Submitted 2025-10-09 · Read on arXiv

cs.AI, cs.CV, cs.RO

Submitted: 2025-10-09

Updated: 2026-10-16

Comments: Accepted to ECCV 2026. 22 pages, 12 figures, code: https://github.com/UWMILab/UniWM

Code: https://github.com/UWMILab/UniWM

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• World Model — 2× in this paper · explained in 7 episodes
• Adaptability — 1× in this paper · explained in 1 episode
• Benchmark — 1× in this paper · explained in 2 episodes

Sources

• Cosmos World Foundation Model Platform for Physical AI
• Walk through Paintings: Egocentric World Models from Internet Priors
• Qwen2.5-VL Technical Report
• Back to the Features: DINO as a Foundation for Video World Models
• Revisiting Feature Prediction for Learning Visual Representations from Video
• VaViM and VaVAM: Autonomous Driving through Video Generative Modeling
• WorldVLA: Towards Autoregressive Action World Model
• Learning to Explore using Active Neural SLAM
• SHIELD: LLM-Driven Schema Induction for Predictive Analytics in EV Battery Supply Chain Disruptions
• ANOLE: An Open, Autoregressive, Native Large Multimodal Models for Interleaved Image-Text Generation
• Driver-WM: A Driver-Centric Traffic-Conditioned Latent World Model for In-Cabin Dynamics Rollout
• Agentic World Modeling: Foundations, Capabilities, Laws, and Beyond
• HA-VLN 2.0: An Open Benchmark and Leaderboard for Human-Aware Navigation in Discrete and Continuous Environments with Dynamic Multi-Human Interactions
• Language-Conditioned World Modeling for Visual Navigation
• GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers
• DreamSim: Learning New Dimensions of Human Visual Similarity using Synthetic Data
• Embodied AI Agents: Modeling the World
• World Models
• Dream to Control: Learning Behaviors by Latent Imagination
• Mastering Atari with Discrete World Models

Related papers

• MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
• Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
• Agents' Last Exam
• The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
• MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
• Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems