⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

HERMES: A Unified Self-Driving World Model for Simultaneous 3D Scene Understanding and Generation

ICCV 25 2025 66.7 method, application

TLDR

A unified driving world model that integrates 3D scene understanding and future scene generation using BEV and LLM-based world queries.

Reasoning

The paper presents a novel integration of scene understanding and generation in a driving world model, achieving strong empirical results on nuScenes. However, it is limited to driving scenarios and does not address interactive or reinforcement learning aspects.

Read-first score

Read-first score 66.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 39.

Recency 8%
86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Reproducibility 25%
85

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=True; dataset=False; markers=code,dataset,github

Methodology quality 25%
60

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=dataset,metric

Topical relevance 42%
55.7

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Field roles

FrontierReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 283.

Keyword Scores

world model
9
generative world model
8
world dynamics prediction
7
video world model
6
world simulator
4
model-based reinforcement learning world model
3
interactive world model
2

Deep Analysis

Innovations

  • Unified framework for simultaneous 3D scene understanding and generation in driving world models
  • Use of Bird's-Eye View (BEV) representation to consolidate multi-view spatial information while preserving geometric relationships
  • Introduction of world queries that incorporate world knowledge into BEV features via causal attention in a Large Language Model

Methodology

HERMES is a unified driving world model that integrates 3D scene understanding and future scene generation. It uses Bird's-Eye View (BEV) representation to consolidate multi-view spatial information and introduces world queries that incorporate world knowledge into BEV features via causal attention in a Large Language Model. The model is evaluated on nuScenes and OmniDrive-nuScenes datasets.

Key Results

HERMES achieves state-of-the-art performance, reducing generation error by 32.4% and improving understanding metrics such as CIDEr by 8.0%.

Resources

PaperCodeProject

Tags