Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Behavioral Analysis of Vision-and-Language Navigation Agents
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 |
|---|---|---|---|
| Give an explicit stop instruction instead of an implicit stopping instruction | HAMT · RxR stop interventions | stop probability: 0.72 vs 0.65, effect: 0.07 | Helped |
| Give an implicit or explicit stop instruction instead of a one-step-ahead instruction | HAMT · RxR stop interventions | stop probability: effect 0.36 | Helped |
| Add an instruction to turn and move in the target direction | HAMT · RxR directional interventions | probability mass in target direction: left (0.38), right (0.44), forward (0.16), backward (0.16), back left (0.43), back right (0.48) | Helped |
| Add an instruction to walk toward a referenced object | HAMT · RxR object-seeking interventions | probability within 15 degrees of object heading: 0.069 | Helped |
| Add an instruction to walk toward a neighboring room type | HAMT · RxR 1-hop room-seeking | delta geodesic distance toward target room: 0.26 | Helped |
| Add room-seeking instructions for rooms several steps away | HAMT · RxR k-hop room-seeking | distance to target room: ≤-0.1 for 3,4, and 5 hops | Helped |
| Use HAMT rather than EnvDrop | VLN agents · RxR skill-specific interventions | average skill score: 31.12 -> 38.50 | Helped |
| Use EnvDrop (ViL CLIP) rather than EnvDrop | VLN agents · RxR skill-specific interventions | Object score: 11.06 -> 12.83 | Helped |
| Use HAMT rather than EnvDrop (ViL CLIP) | VLN agents · RxR skill-specific interventions | Turn score: 27.45 -> 43.74 | Helped |
| Add teacher forcing along the partial ground-truth trajectory (ablation) | HAMT · RxR | SR: 42.49 -> 52.94 | Helped |
Closest papers by meaning
- Curriculum Learning for Vision-and-Language Navigation (NeurIPS 2021 · similarity 0.91)
- Generative Language-Grounded Policy in Vision-and-Language Navigation with Bayes' Rule (ICLR 2021 · similarity 0.89)
- Do Visual Imaginations Improve Vision-and-Language Navigation Agents? (CVPR 2025 · similarity 0.89)
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.88)
- General Scene Adaptation for Vision-and-Language Navigation (ICLR 2025 · similarity 0.88)
- GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning (NeurIPS 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.