Research map / Vision-language and multimodal / Visual question answering
SAFE-EQA: Semantic-Aware Efficient Exploration for Embodied Question Answering
Yijie Tang, Ke Xia, Jiazhao Zhang, Zhinan Yu, Zhiyuan Yu, Dezun Dong, Renjiao Yi, Chenyang Zhu, Kai Xu
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 |
|---|---|---|---|
| use semantic-aware TSP instead of distance-only TSP (ablation) | SAFE-EQA · OpenEQA | LLM-Match SPL: 45.7 -> 53.9 | Improved |
| use semantic-aware TSP instead of distance-only TSP (ablation) | SAFE-EQA · OpenEQA | tokens per scene: 14376.9 -> 10585.8 | Improved |
| add frontier-triggered priority assessment (ablation) | SAFE-EQA · EXPRESS-Bench long trajectories (>25 m) | trajectory length (m): 39.4 -> 32.7 | Improved |
| add frontier-triggered priority assessment (ablation) | SAFE-EQA · EXPRESS-Bench long trajectories (>25 m) | tokens: 35921.3 -> 38753.2 | Hurt |
| SAFE-EQA vs 3D-Mem, multi-step TSP planning with hierarchical answering | SAFE-EQA · OpenEQA | tokens per scene: 19488.7 -> 10585.8 | Improved |
| SAFE-EQA vs 3D-Mem, multi-step TSP planning with hierarchical answering | SAFE-EQA · EXPRESS-Bench | tokens per scene: 23270.1 -> 14991.5 | Improved |
| SAFE-EQA vs 3D-Mem, multi-step TSP planning with hierarchical answering | SAFE-EQA · HM-EQA | tokens per scene: 22461.7 -> 16214.1 | Improved |
| SAFE-EQA vs 3D-Mem on exploration efficiency | SAFE-EQA · OpenEQA | LLM-Match SPL: 42.0 -> 53.9 | Improved |
| SAFE-EQA vs 3D-Mem on answering accuracy | SAFE-EQA · EXPRESS-Bench | LLM-Match: 67.5 -> 63.3 | Hurt |
| SAFE-EQA vs 3D-Mem on answering accuracy | SAFE-EQA · OpenEQA | LLM-Match: 52.6 -> 56.4 | Improved |
| SAFE-EQA vs 3D-Mem on answering accuracy | SAFE-EQA · HM-EQA† | LLM-Match: 59.0 -> 59.2 | No clear effect |
Closest papers by meaning
- Beyond the Destination: A Novel Benchmark for Exploration-Aware Embodied Question Answering (ICCV 2025 · similarity 0.95)
- Predict Before You Explore: Predictive Planning with Specialized Memory for Embodied Question Answering (CVPR 2026 · similarity 0.92)
- OpenEQA: Embodied Question Answering in the Era of Foundation Models (CVPR 2024 · similarity 0.90)
- Multi-modal Situated Reasoning in 3D Scenes (NeurIPS 2024 · similarity 0.88)
- Embodied Scene Understanding for Vision Language Models via MetaVQA (CVPR 2025 · similarity 0.88)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 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.
