Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Soft Expert Reward Learning for Vision-and-Language Navigation
Hu Wang, Qi Wu, Chunhua Shen
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 Soft Expert Distillation module (ablation) | EnvDrop-based agent · R2R Val Unseen | SR: 0.49 -> 0.52 | Improved |
| add Soft Expert Distillation module (ablation) | EnvDrop-based agent · R2R Val Unseen | SPL: 0.45 -> 0.48 | Improved |
| add Self Perceiving module (ablation) | EnvDrop-based agent · R2R Val Unseen | SR: 0.49 -> 0.53 | Improved |
| add Self Perceiving module (ablation) | EnvDrop-based agent · R2R Val Unseen | SPL: 0.45 -> 0.46 | Improved |
| add SP module on top of SED (ablation) | SERL · R2R Val Unseen | SR: 0.52 -> 0.56 | Improved |
| add SED module on top of SP (ablation) | SERL · R2R Val Unseen | SR: 0.53 -> 0.56 | Improved |
| add both SED and SP modules (ablation) | EnvDrop-based agent · R2R Val Seen | SR: 0.66 -> 0.69 | Improved |
| add both SED and SP modules (ablation) | EnvDrop-based agent · R2R Val Unseen | NE: 5.49 -> 4.74 | Improved |
| use beam search at inference (ablation) | SERL · R2R Val Unseen | SR: 0.56 -> 0.71 | Improved |
| use beam search at inference (ablation) | SERL · R2R Val Unseen | SPL: 0.48 -> 0.02 | Hurt |
| use beam search at inference (ablation) | SERL · R2R Val Seen | SR: 0.69 -> 0.77 | Improved |
| vary reward trade-off weights alpha and beta (ablation) | SERL · R2R Val Unseen | SR | No clear effect |
Closest papers by meaning
- Vision-Language Navigation with Energy-Based Policy (NeurIPS 2024 · similarity 0.88)
- Code as Reward: Empowering Reinforcement Learning with VLMs (ICML 2024 · similarity 0.86)
- FuRL: Visual-Language Models as Fuzzy Rewards for Reinforcement Learning (ICML 2024 · similarity 0.86)
- Selective Expert Guidance for Effective and Diverse Exploration in Reinforcement Learning of LLMs (ICLR 2026 · similarity 0.86)
- Vision-Language Models are Zero-Shot Reward Models for Reinforcement Learning (ICLR 2024 · similarity 0.85)
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · similarity 0.85)
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.
