Research map / Vision-language and multimodal / Robot action and navigation
Simulation to Rules: A Dual-VLM Framework for Formal Visual Planning
Yilun Hao, Yongchao Chen, Chuchu Fan, Yang Zhang
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 prescreening (ablation) | VLMFP · 6 grid worlds avg (seen appearance) | success rate (%): 47.5 -> 70.0 | Improved |
| add simulation feedback (ablation) | VLMFP · 6 grid worlds avg (seen appearance) | success rate (%): 61.1 -> 70.0 | Improved |
| add PDDL updating (ablation) | VLMFP · 6 grid worlds avg (seen appearance) | success rate (%): 30.7 -> 70.0 | Improved |
| add prescreening on Frozenlake (ablation) | VLMFP · Frozenlake (seen appearance) | success rate (%): 94.3 -> 95.2 | No clear effect |
| add feedback on Frozenlake (ablation) | VLMFP · Frozenlake (seen appearance) | success rate (%): 95.5 -> 95.2 | Hurt |
| use VLMFP instead of CodePDDL (GPT-4o) | GPT-4o · 6 grid worlds avg (seen appearance) | success rate (%): 30.7 -> 70.0 | Improved |
| use VLMFP instead of CodePDDL on unseen appearance | GPT-4o · 6 grid worlds avg (unseen appearance) | success rate (%): 32.3 -> 54.1 | Improved |
| use VLMFP instead of CodePDDL on Sokoban | GPT-4o · Sokoban (unseen appearance) | success rate (%): 0.4 -> 25.1 | Improved |
| use GPT-5 instead of GPT-4o for direct planning | Direct · 6 grid worlds avg (seen appearance) | success rate (%): 1.3 -> 11.8 | Improved |
| use CoT instead of direct prompting with GPT-5 | GPT-5 · 6 grid worlds avg (seen appearance) | success rate (%): 11.8 -> 12.7 | No clear effect |
| test SimVLM on unseen instead of seen appearance (task description) | SimVLM · 6 grid worlds avg | string matching rate (%): 95.5 -> 82.6 | Hurt |
| test SimVLM on unseen instead of seen appearance (all outputs) | SimVLM · 6 grid worlds avg | string matching rate (%): 87.3 -> 86.0 | No clear effect |
Closest papers by meaning
- Visual Planning: Let's Think Only with Images (ICLR 2026 · similarity 0.90)
- Evaluating Vision-Language Models as Evaluators in Path Planning (CVPR 2025 · similarity 0.89)
- World-aware Planning Narratives Enhance Large Vision-Language Model Planner (NeurIPS 2025 · similarity 0.89)
- From Representational Complementarity to Dual Systems: Synergizing VLM and Vision-Only Backbones for End-to-End Driving (NeurIPS 2026 · similarity 0.88)
- Video Language Planning (ICLR 2024 · similarity 0.88)
- APIVOT: Adaptive Planning with Interleaved Vision-Language Thoughts (NeurIPS 2026 · similarity 0.88)
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.
