Research map / Vision-language and multimodal / Robot action and navigation
Contrastive Instruction-Trajectory Learning for Vision-Language Navigation
Read the paperCheck 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 CITL to RecBERT initialized from PREVALENT | RecBERT (init PREVALENT) · R2R Test Unseen | SPL: 57 -> 59 | Helped |
| Add CITL to RecBERT initialized from PREVALENT | RecBERT (init PREVALENT) · R4R Val Unseen | SPL: 31.7 -> 35.1 | Helped |
| Add CITL to RecBERT | RecBERT (init PREVALENT) · RxR monolingual unseen validation | SPL: 39.3 -> 40.7 | Helped |
| Add pair mining to multi-pair InfoNCE with a memory bank (ablation) | Trajectory contrastive loss · R2R Val Unseen | SPL: 52.62 -> 54.10 | Helped |
| Add a memory bank to circle loss without pair mining (ablation) | Trajectory contrastive loss · R2R Val Unseen | SPL: 53.35 -> 53.14 | Hurt |
| Add pair mining to circle loss without a memory bank (ablation) | Trajectory contrastive loss · R2R Val Unseen | SPL: 53.35 -> 54.54 | Helped |
| Use the full pairwise sample-reweighting mechanism instead of multi-pair InfoNCE with a memory bank (ablation) | Trajectory contrastive loss · R2R Val Unseen | SPL: 52.62 -> 55.47 | Helped |
| Add coarse trajectory contrastive loss (ablation) | RecBERT baseline · R2R Val Unseen | SPL: 52.90 -> 55.47 | Helped |
| Add coarse instruction contrastive loss (ablation) | RecBERT baseline · R2R Val Unseen | SPL: 52.90 -> 55.17 | Helped |
| Add fine-grained sub-instruction contrastive loss (ablation) | RecBERT baseline · R2R Val Unseen | SPL: 52.90 -> 53.58 | Helped |
| Combine coarse trajectory, coarse instruction, and fine sub-instruction contrastive losses (ablation) | RecBERT baseline · R2R Val Unseen | SPL: 52.90 -> 55.83 | Helped |
Closest papers by meaning
- Towards Learning a Generic Agent for Vision-and-Language Navigation via Pre-Training (CVPR 2020 · similarity 0.91)
- A New Path: Scaling Vision-and-Language Navigation With Synthetic Instructions and Imitation Learning (CVPR 2023 · similarity 0.90)
- Counterfactual Cycle-Consistent Learning for Instruction Following and Generation in Vision-Language Navigation (CVPR 2022 · similarity 0.90)
- Progress-Think: Semantic Progress Reasoning for Vision-Language Navigation (CVPR 2026 · similarity 0.90)
- Instruction Decomposition and Action Alignment for Vision-Language Navigation (ICML 2026 · similarity 0.89)
- Towards Long-Horizon Vision-Language Navigation: Platform, Benchmark and Method (CVPR 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.