Research map / Vision-language and multimodal / Robot action and navigation
Look Before You Leap: Bridging Model-Free and Model-Based Reinforcement Learning for Planned-Ahead Vision-and-Language Navigation
Xin Wang, Wenhan Xiong, Hongmin Wang, William Yang Wang
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 look-ahead module on top of model-free RL | Model-free RL · R2R Val Unseen | success rate (%): 21.5 -> 24.6 | Improved |
| add look-ahead module on top of model-free RL | Model-free RL · R2R Val Seen | success rate (%): 41.9 -> 42.9 | Improved |
| add look-ahead module on top of model-free RL | Model-free RL · R2R Test (unseen) | success rate (%): 23.1 -> 25.3 | Improved |
| add look-ahead module on top of model-free RL | Model-free RL · R2R Val Unseen | navigation error (m): 7.88 -> 7.65 | Improved |
| switch cross-entropy training to model-free RL | XE · R2R Val Unseen | success rate (%): 21.3 -> 21.5 | No clear effect |
| switch cross-entropy training to model-free RL | XE · R2R Val Seen | success rate (%): 40.2 -> 41.9 | Improved |
| use per-step discounted reward instead of global distance reward (ablation) | Model-free RL · R2R Val Seen | success rate (%): 35.5 -> 40.5 | Improved |
| use per-step discounted reward instead of binary success reward (ablation) | Model-free RL · R2R Val Seen | success rate (%): 37.8 -> 40.5 | Improved |
| add success reward to discounted reward (ablation) | Model-free RL · R2R Val Seen | success rate (%): 40.5 -> 41.9 | Improved |
| add success reward to discounted reward (ablation) | Model-free RL · R2R Val Unseen | success rate (%): 20.4 -> 21.5 | Improved |
| add success reward to discounted reward (ablation) | Model-free RL · R2R Val Unseen | navigation error (m): 7.74 -> 7.88 | Hurt |
| use per-step discounted reward instead of global distance reward (ablation) | Model-free RL · R2R Val Unseen | success rate (%): 19.0 -> 20.4 | Improved |
Closest papers by meaning
- Ground Slow, Move Fast: A Dual-System Foundation Model for Generalizable Vision-Language Navigation (ICLR 2026 · similarity 0.90)
- Toward a Causally-Inspired Evolving World Model for Vision-and-Language Navigation in Continuous Environments (TPAMI 2026 · similarity 0.89)
- Towards Practical World Model-based Reinforcement Learning for Vision-Language-Action Models (ICML 2026 · similarity 0.89)
- Test-Time Adaptation for Online Vision-Language Navigation with Feedback-based Reinforcement Learning (ICML 2025 · similarity 0.88)
- Bridging Environments and Language with Rendering Functions and Vision-Language Models (ICML 2024 · similarity 0.88)
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · 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.
