Research map / Vision-language and multimodal / Robot action and navigation
Counterfactual Vision-and-Language Navigation: Unravelling the Unseen
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 counterfactual objective with optimized interventions to imitation-learning agent | IL agent · R2R val-unseen | success rate (%): 42.6 -> 46.4 | Helped |
| add counterfactual objective to IL+RL agent | IL+RL agent · R2R val-unseen | success rate (%): 44.9 -> 49.4 | Helped |
| add counterfactual objective to IL+RL+Aug agent | IL+RL+Aug agent · R2R val-unseen | success rate (%): 50.3 -> 53.5 | Helped |
| add counterfactual objective to IL agent | IL agent · R2R val-seen | success rate (%): 50.2 -> 48.9 | Hurt |
| add counterfactual objective to IL+RL agent | IL+RL agent · R2R val-seen | success rate (%): 55.8 -> 53.6 | Hurt |
| add counterfactual objective to IL+RL+Aug agent | IL+RL+Aug agent · R2R val-seen | success rate (%): 62.5 -> 68.2 | Helped |
| optimize intervention via Eq. 11 (posterior) instead of sampling from prior (ablation) | IL agent with counterfactuals · R2R val-unseen | success rate (%): 45.5 -> 46.4 | Helped |
| optimize intervention instead of sampling from prior (ablation) | IL+RL+Aug agent with counterfactuals · R2R val-unseen | success rate (%): 52.4 -> 53.5 | Helped |
| sample counterfactual intervention from the prior only (ablation) | IL agent · R2R val-unseen | success rate (%): 42.6 -> 45.5 | Helped |
| add counterfactual objective to GRU navigation policy | GRU policy · EQA unseen T-30 | distance to target dT (m): 3.99 -> 3.88 | Helped |
| add counterfactual objective to GRU navigation policy | GRU policy · EQA unseen T-50 | distance to target dT (m): 8.73 -> 8.46 | Helped |
| increase counterfactual weight alpha/(1-alpha) beyond optimum (ablation) | IL agent · R2R val-unseen | success rate (%) | Mixed |
Closest papers by meaning
- Counterfactual Cycle-Consistent Learning for Instruction Following and Generation in Vision-Language Navigation (CVPR 2022 · similarity 0.90)
- Curriculum Learning for Vision-and-Language Navigation (NeurIPS 2021 · similarity 0.90)
- CAVEN: An Embodied Conversational Agent for Efficient Audio-Visual Navigation in Noisy Environments (AAAI 2024 · similarity 0.87)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 2025 · similarity 0.87)
- Human-Aware Vision-and-Language Navigation: Bridging Simulation to Reality with Dynamic Human Interactions (NeurIPS 2024 · similarity 0.87)
- General Scene Adaptation for Vision-and-Language Navigation (ICLR 2025 · similarity 0.87)
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.