Research map / Vision-language and multimodal / Robot action and navigation
Counterfactual Cycle-Consistent Learning for Instruction Following and Generation in 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 |
|---|---|---|---|
| Apply CCC to E-Dropout | E-Dropout · R2R test unseen | SR: 48.0 -> 52.2 | Helped |
| Apply CCC to Active Perception | Active Perception · R2R test unseen | SR: 55.7 -> 60.6 | Helped |
| Apply CCC to SSM | SSM · R2R test unseen | SR: 57.3 -> 62.2 | Helped |
| Jointly train the speaker and follower with CCC | R2R instruction generation · R2R val unseen | SPICE: 0.188 -> 0.214 | Helped |
| Use both labeled cycle directions instead of the independent baseline (ablation) | Speaker-Follower · R2R val unseen | SR: 29.9 -> 35.1 | Helped |
| Use only the instruction-to-route cycle (ablation) | Speaker-Follower · R2R val unseen | SR: 29.9 -> 28.6 | Hurt |
| Add unlabeled sampled paths to the two labeled cycle losses (ablation) | Speaker-Follower · R2R val unseen | SR: 35.1 -> 37.7 | Helped |
| Use a reference environment when creating counterfactual scenes (ablation) | Counterfactual creator · R2R val unseen | SR: 29.9 -> 41.7 | Helped |
| Add counterfactual samples to the cycle-consistent learning system (ablation) | Speaker-Follower · R2R val unseen | SR: 37.7 -> 43.5 | Helped |
| Use both L2 and adversarial losses to train the creator (ablation) | Counterfactual creator · R2R val unseen | SR: 38.3 -> 43.5 | Helped |
| Add the adversarial loss to the creator's L2 loss (ablation) | Counterfactual creator · R2R val unseen | SR: 32.7 -> 43.5 | Helped |
Closest papers by meaning
- Speaker-Follower Models for Vision-and-Language Navigation (NeurIPS 2018 · similarity 0.91)
- Instruction Decomposition and Action Alignment for Vision-Language Navigation (ICML 2026 · similarity 0.90)
- Counterfactual Vision-and-Language Navigation: Unravelling the Unseen (NeurIPS 2020 · similarity 0.90)
- Contrastive Instruction-Trajectory Learning for Vision-Language Navigation (AAAI 2022 · similarity 0.90)
- A New Path: Scaling Vision-and-Language Navigation With Synthetic Instructions and Imitation Learning (CVPR 2023 · similarity 0.90)
- General Scene Adaptation for Vision-and-Language Navigation (ICLR 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.