HERMES++: Toward a Unified Driving World Model for 3D Scene Understanding and Generation

2026-04-30

AI Summary

This work addresses the disconnect between 3D scene understanding and future geometric prediction in existing driving world models, as well as the inability of large language models (LLMs) to effectively model physical dynamics despite their strong semantic reasoning capabilities. To bridge this gap, the authors propose the first unified framework that integrates multi-view spatial information through a bird’s-eye-view (BEV) representation, enhances world queries with LLM-derived prior knowledge, and introduces a temporal linking mechanism from present to future states alongside a joint geometric optimization strategy—combining explicit geometric constraints with implicit latent regularization. The method significantly outperforms specialized models across multiple benchmarks, achieving state-of-the-art performance in both future point cloud prediction and 3D scene understanding, thereby effectively reconciling semantic reasoning with geometric modeling.

Abstract

Driving world models serve as a pivotal technology for autonomous driving by simulating environmental dynamics. However, existing approaches predominantly focus on future scene generation, often overlooking comprehensive 3D scene understanding. Conversely, while Large Language Models (LLMs) demonstrate impressive reasoning capabilities, they lack the capacity to predict future geometric evolution, creating a significant disparity between semantic interpretation and physical simulation. To bridge this gap, we propose HERMES++, a unified driving world model that integrates 3D scene understanding and future geometry prediction within a single framework. Our approach addresses the distinct requirements of these tasks through synergistic designs. First, a BEV representation consolidates multi-view spatial information into a structure compatible with LLMs. Second, we introduce LLM-enhanced world queries to facilitate knowledge transfer from the understanding branch. Third, a Current-to-Future Link is designed to bridge the temporal gap, conditioning geometric evolution on semantic context. Finally, to enforce structural integrity, we employ a Joint Geometric Optimization strategy that integrates explicit geometric constraints with implicit latent regularization to align internal representations with geometry-aware priors. Extensive evaluations on multiple benchmarks validate the effectiveness of our method. HERMES++ achieves strong performance, outperforming specialist approaches in both future point cloud prediction and 3D scene understanding tasks. The model and code will be publicly released at https://github.com/H-EmbodVis/HERMESV2.

Problem

• driving world model
• 3D scene understanding
• future geometry prediction
• semantic-physical gap
• autonomous driving

Innovation

• unified world model
• 3D scene understanding
• future geometry prediction
• LLM-enhanced queries
• joint geometric optimization