Research map / Reinforcement learning / Visual and representation RL
VPA-WM: Vision-Priors-Aligned World Models for Robust Visual Reinforcement Learning
Xu Zhang, Sicong Liu, Liwei Guo, Chenjuan Guo, Bin Yang, Yang Shu
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 |
|---|---|---|---|
| add control-oriented alignment fine-tuning of PVM (ablation) | DA-WM · Distracted DMC Walker Run | episode return: 424 -> 572 | Improved |
| add instant action prediction (ablation) | DA-WM · Distracted DMC Walker Run | episode return: 219 -> 572 | Improved |
| add short-horizon return estimation (ablation) | DA-WM · Distracted DMC Cheetah Run | episode return: 402 -> 804 | Improved |
| add spatially grounded guidance (ablation) | DA-WM · Distracted DMC Cheetah Run | episode return: 154 -> 804 | Improved |
| add experience query integration (ablation) | DA-WM · Distracted DMC Walker Run | episode return: 478 -> 572 | Improved |
| overlay DINO masks on input instead of STCI (ablation) | DA-WM · Distracted DMC Walker Run | episode return: 572 -> 146 | Hurt |
| replace world model encoder with DINO (ablation) | DA-WM · Distracted DMC Cheetah Run | episode return: 804 -> 349 | Hurt |
| use DA-WM instead of Dreamer under distractions | Dreamer · Distracted DMC Cheetah Run | episode return: 104 -> 804 | Improved |
| use DA-WM instead of Dreamer on clean Standard DMC | DA-WM · Standard DMC Cheetah Run | episode return: 892 -> 789 | No clear effect |
| swap DINOv2 for CLIP as PVM (CA-WM) | VPA-WM · Distracted DMC Cheetah Run | episode return: 804 -> 699 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- RePo — RePo: Resilient Model-Based Reinforcement Learning by Regularizing Posterior Predictability (NeurIPS 2023) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning Latent Dynamic Robust Representations for World Models (ICML 2024 · similarity 0.88)
- DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning (ICML 2025 · similarity 0.88)
- COVR: Collaborative Optimization of VLMs and RL Agent for Visual-Based Control (AAAI 2026 · similarity 0.88)
- SyncWorld: Visual Calibration Enables World Models as Zero-Shot Simulators (NeurIPS 2026 · similarity 0.87)
- Reinforcing VLAs in Task-Agnostic World Models (NeurIPS 2026 · similarity 0.87)
- Learning Visual Feature-Based World Models via Residual Latent Action (NeurIPS 2026 · 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.
