Research map / Vision-language and multimodal / Robot action and navigation
AerialVLA: A Vision-Language-Action Model for Aerial Navigation with Online Dialogue
Jinyu Chen, Hongyu Li, Zongheng Tang, Xiaoduo Li, Wenjun Wu, Si Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 OTDA data augmentation (ablation) | AerialVLA baseline · AVDH seen validation | SPL: 3.1 -> 13.0 | Improved |
| add OTDA data augmentation (ablation) | AerialVLA baseline · AVDH unseen validation | SPL: 5.5 -> 17.3 | Improved |
| replace video features with global map of history (ablation) | AerialVLA with OTDA · AVDH seen validation | SPL: 13.0 -> 21.1 | Improved |
| replace video features with global map of history (ablation) | AerialVLA with OTDA · AVDH unseen validation | SPL: 17.3 -> 19.5 | Improved |
| add Spatial-Temporal Attention (ablation) | AerialVLA with OTDA and global map · AVDH unseen validation | SPL: 19.5 -> 26.6 | Improved |
| add Spatial-Temporal Attention (ablation) | AerialVLA with OTDA and global map · AVDH seen validation | SPL: 21.1 -> 25.4 | Improved |
| use original-to-generated data ratio 1:2 instead of 1:1 (ablation) | OTDA · AVDH unseen validation | SPL: 14.4 -> 26.6 | Improved |
| increase generated data ratio from 1:2 to 1:3 (ablation) | OTDA · AVDH unseen validation | SPL: 26.6 -> 20.5 | Hurt |
| use STA distances 1,2,3 instead of higher 1,3,5 (ablation) | STA · AVDH unseen validation | SPL: 24.0 -> 26.6 | Improved |
| use STA distances 1,2,3 instead of lower 1,1,1 (ablation) | STA · AVDH unseen validation | SPL: 23.0 -> 26.6 | Improved |
| add PNaQ proactive questioning (online dialogue) instead of pre-recorded dialogues | AerialVLA · UNOD unseen validation | SPL: 13.7 -> 21.6 | Improved |
| use AerialVLA instead of HAA-T on long-dialogue benchmark | AVDH-Full unseen testing · AVDH-Full unseen testing | SPL: 4.3 -> 17.1 | Improved |
Closest papers by meaning
- Towards Realistic UAV Vision-Language Navigation: Platform, Benchmark, and Methodology (ICLR 2025 · similarity 0.90)
- Parse, Search, and Confirmation: Training-Free Aerial Vision-and-Dialog Navigation with Chain-of-Thought Reasoning and Structured Spatial Memory (CVPR 2026 · similarity 0.90)
- WorldVLN: Autoregressive World Action Model for Aerial Vision-Language Navigation (NeurIPS 2026 · similarity 0.89)
- SatNav: A Scalable Benchmark for Long-Horizon UAV Vision-Language Navigation from Satellite Imagery (NeurIPS 2026 · similarity 0.87)
- UAV-Flow Colosseo: A Real-World Benchmark for Flying-on-a-Word UAV Imitation Learning (NeurIPS 2025 · similarity 0.87)
- Learning Fine-Grained Alignment for Aerial Vision-Dialog Navigation (AAAI 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.
