Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Cloning Deterministic Worlds: The Critical Role of Latent Geometry in Long-Horizon World Models
Zaishuo Xia, Yukuan Lu, Xinyi Li, Yifan Xu, Yubei Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use ground-truth state (oracle) instead of VAE latent (ablation) | world model · deterministic 3D navigation | frame-wise MSE: near-zero error vs rapid accumulation | Improved |
| add temporal contrastive geometric regularization (GRWM) | Diffusion Forcing, Video Diffusion, Standard Diffusion world models · M3x3-DET, M9x9-DET, MC-DET | frame-wise MSE over 63 steps | Improved |
| add GRWM regularization | VAE-WM autoencoder · M3×3-DET latent probing | probe MSE: 0.082 -> 0.031 | Improved |
| add GRWM regularization | VAE-WM autoencoder · M9×9-DET latent probing | probe MSE: 0.106 -> 0.058 | Improved |
| add GRWM regularization | VAE-WM autoencoder · MC-DET latent probing | probe MSE: 0.137 -> 0.081 | Improved |
| add GRWM regularization | Diffusion Forcing · 10,000-frame rollout M9x9 | rollout coherence | Improved |
| add GRWM regularization | world model latent space · k-means clustering (k=20) | cluster spatial coherence | Improved |
Closest papers by meaning
- Learning 3D Persistent Embodied World Models (NeurIPS 2025 · similarity 0.88)
- Simplifying Latent Dynamics with Softly State-Invariant World Models (NeurIPS 2024 · similarity 0.88)
- Zero-shot World Models via Search in Memory (NeurIPS 2025 · similarity 0.88)
- Dyn-O: Building Structured World Models with Object-Centric Representations (NeurIPS 2025 · similarity 0.87)
- Predictive but Not Plannable: RC-aux for Latent World Models (NeurIPS 2026 · similarity 0.87)
- WALL-E: World Alignment by NeuroSymbolic Learning improves World Model-based LLM Agents (NeurIPS 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
