Research map / Vision-language and multimodal / Multimodal LLMs
What if? Emulative Simulation with World Models for Situated Reasoning
Ruiping Liu, Yufan Chen, Yuheng Zhang, Junwei Zheng, Kunyu Peng, Chengzhi Wu, Chenguang Huang, Di Wen, Jiaming Zhang, Kailun Yang, Rainer Stiefelhagen
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 |
|---|---|---|---|
| fine-tune Wan2.1 on WanderDream (LoRA) instead of prompt-extended Wan2.2 | world model · Real-world test set | FVD: 41.65 -> 27.49 | Improved |
| fine-tune Wan2.1 on WanderDream instead of prompt-extended Wan2.2 | sequential framework · Real-world test set | QA accuracy: 38.8 -> 43.0 | Improved |
| add MindJourney step-wise imagination | MLLM reasoning · Real-world test set | QA accuracy: 38.5 -> 31.9 | Hurt |
| fine-tune (LoRA) Wan2.2 instead of prompt extension (ablation) | Wan2.2 · WanderDream-Gen (HM3D val) | FVD: 17.62 -> 6.19 | Improved |
| fine-tune (LoRA) Wan2.2 instead of prompt extension (ablation) | Wan2.2 · WanderDream-Gen (ScanNet++ val) | FVD: 18.73 -> 9.67 | Improved |
| add imagined video frames to start-state reasoning (ablation) | Qwen3-VL · WanderDream-QA ScanNet++ start-state | correctness (%): 32.5 -> 29.7 | Hurt |
| add imagined video (CogVideoX1.5 SFT) to end-state reasoning (ablation) | Qwen3-VL · WanderDream-QA ScanNet++ end-state | correctness (%): 45.5 -> 47.5 | Improved |
| add imagined video (CogVideoX1.5 SFT) to end-state reasoning (ablation) | Qwen3-VL · WanderDream-QA HM3D end-state | correctness (%): 41.8 -> 44.0 | Improved |
| add imagined video (Wan2.1 LoRA) to path reasoning (ablation) | Qwen3-VL · WanderDream-QA HM3D path | correctness (%): 53.2 -> 53.4 | No clear effect |
| add ground-truth intermediate frames to end-state ESR reasoning (ablation) | Qwen3-VL · HM3D per-class QA | correctness (%): 26.0 -> 38.7 | Improved |
| add imagination in ScanNet++ path questions where anchor is visible (ablation) | Qwen3-VL · ScanNet++ path questions | correctness (%) | No clear effect |
| use better video generation quality for imagination | sequential framework · WanderDream-QA | correctness (%) | Improved |
Closest papers by meaning
- Thinking with Imagination: Agentic Visual Spatial Reasoning with World Simulators (NeurIPS 2026 · similarity 0.90)
- MindJourney: Test-Time Scaling with World Models for Spatial Reasoning (NeurIPS 2025 · similarity 0.88)
- Active Exploring like a Pigeon: Reinforcing Spatial Reasoning via Agentic Vision-Language Models (ICML 2026 · similarity 0.88)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.87)
- FantasyVLN: Unified Multimodal Chain-of-Thought Reasoning for Vision-and-Language Navigation (CVPR 2026 · similarity 0.87)
- EvoEmpirBench: Dynamic Spatial Reasoning with Agent-ExpVer (AAAI 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.
