ATLAS: aligned transport objective preserves latent geometry for reliable world-model planning
Ke Fang · hf · 2026-10-01
This work identifies a blind spot in latent world models: regularizing the latent marginal alone doesn't preserve the state-to-state relationships used for action selection, weakening planning-relevant novelty structure.
- ATLAS transfers normalized pairwise structure from an informative encoder representation to the planning latent and calibrates the marginal via 1D Wasserstein-2 embedding matching (WEMReg).
- Analysis shows relational preservation and marginal calibration are non-redundant constraints, linking planning stability to relational distortion, scale mismatch, and prediction error.
- Instantiated in LeWM, ATLAS improves goal-reaching across PushT, TwoRoom, and OGBench-Cube, with the largest gain on higher-novelty TwoRoom episodes; diagnostics show stronger novelty structure and lower multi-step prediction error.
Code: anonymous.4open.science/r/atlas-world-model-72C4
More from Research
- Google Research: generative UI lets teachers build learning simulations, rated 8/10 — dl_weekly · 2026-10-01
- A 1-cent verifier catches 61% of AI agents falsely claiming task completion, paper finds — alex_verem · 2026-10-01
- How Tri Dao's FlashAttention became a cornerstone of modern LLM training — thisdudelikesAI · 2026-10-01
- Looped Transformers: when is recurrent depth worth the extra compute? — ArchitectingAI · 2026-10-01
- ID Balancing applies PID control to stabilize MoE training at 256x sparsity — teortaxesTex · 2026-10-01
- AI-Newton derives F=ma, gravity and energy conservation from noisy data with zero prior physics — CurieuxExplorer · 2026-10-01