Research map / Vision-language and multimodal / Robot action and navigation
Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 RCM cross-modal matching and reasoning | VLN agent · R2R test | SPL: 28 -> 35 | Helped |
| Add SIL using the agent's past good trajectories on training environments | RCM · R2R test | SPL: 35 -> 38 | Helped |
| Explore unseen environments with SIL and imitate selected past trajectories | RCM · R2R test | SR: 43.1 -> 60.5 | Helped |
| Add intrinsic cycle-reconstruction reward (ablation) | RCM · R2R unseen validation | SR: 40.6 -> 42.5 | Helped |
| Add intrinsic cycle-reconstruction reward (ablation) | RCM · R2R seen validation | SR: 67.6 -> 67.4 | Hurt |
| Add extrinsic reward and reinforcement learning rather than pure supervised learning (ablation) | RCM · R2R unseen validation | SR: 37.7 -> 42.5 | Helped |
| Use cross-modal reasoning rather than attention-based seq2seq (ablation) | RCM · R2R unseen validation | SR: 35.7 -> 42.5 | Helped |
| Use SIL to explore unseen environments (ablation) | RCM · R2R unseen validation | SR: 42.5 -> 61.3 | Helped |
| Use dense bottom-up visual features (ablation) | RCM · R2R unseen environments | performance | Hurt |
| Add positional encoding to instructions (ablation) | RCM · R2R | performance | No clear effect |
| Use pretrained ELMo embeddings without fine-tuning (ablation) | RCM · R2R | performance | Hurt |
| Train with intrinsic reward only rather than mixed rewards (ablation) | RCM · R2R | performance | Hurt |
Closest papers by meaning
- Embodied Multi-Modal Agent trained by an LLM from a Parallel TextWorld (CVPR 2024 · similarity 0.89)
- VLMimic: Vision Language Models are Visual Imitation Learner for Fine-grained Actions (NeurIPS 2024 · similarity 0.88)
- Unified Reinforcement and Imitation Learning for Vision-Language Models (NeurIPS 2025 · similarity 0.88)
- Vision-Language Models Create Cross-Modal Task Representations (ICML 2025 · similarity 0.88)
- D3D-VLP: Dynamic 3D Vision-Language-Planning Model for Embodied Grounding and Navigation (CVPR 2026 · similarity 0.88)
- Vision-and-Language Navigation via Causal Learning (CVPR 2024 · 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.