Research map / Vision-language and multimodal / Visual question answering
Beyond the Destination: A Novel Benchmark for Exploration-Aware Embodied Question Answering
Kaixuan Jiang, Yang Liu, Weixing Chen, Jingzhou Luo, Ziliang Chen, Ling Pan, Guanbin Li, Liang Lin
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 frontier-based exploration (ablation) | Fine-EQA · EXPRESS-Bench | C: 38.54 -> 40.55 | Improved |
| add frontier-based exploration (ablation) | Fine-EQA · EXPRESS-Bench | E_path: 12.74 -> 16.22 | Improved |
| add goal-oriented exploration (ablation) | Fine-EQA · EXPRESS-Bench | C: 39.63 -> 40.55 | Improved |
| add goal-oriented exploration (ablation) | Fine-EQA · EXPRESS-Bench | C*: 60.74 -> 63.95 | Improved |
| add goal-oriented exploration (ablation) | Fine-EQA · EXPRESS-Bench | E_path: 14.64 -> 16.22 | Improved |
| use GPT-4o-mini instead of Janus-Pro-7B as VLM (ablation) | Fine-EQA · EXPRESS-Bench | C: 39.88 -> 40.55 | Mixed |
| use GPT-4o-mini instead of Janus-Pro-7B as VLM (ablation) | Fine-EQA · EXPRESS-Bench | E_path: 20.86 -> 16.22 | Mixed |
| use GPT-4o-mini instead of Qwen-Vl-Plus as VLM (ablation) | Fine-EQA · EXPRESS-Bench | C: 40.02 -> 40.55 | Improved |
| replace random exploration with Fine-EQA strategy | exploring agent · EXPRESS-Bench | E_path: 12.06 -> 16.22 | Improved |
| replace frontier-only exploration with Fine-EQA strategy | exploring agent · EXPRESS-Bench | C: 38.60 -> 40.55 | Improved |
| use multiple frames directly instead of captions plus LLM (Socratic) | VLM baseline · EXPRESS-Bench | C*: 62.60 -> 58.37 | Hurt |
| add captioned visual frames to blind LLM DeepSeek-V3 | DeepSeek-V3 · EXPRESS-Bench | C*: 59.15 -> 62.60 | Improved |
Closest papers by meaning
- SAFE-EQA: Semantic-Aware Efficient Exploration for Embodied Question Answering (ECCV 2026 · similarity 0.95)
- OpenEQA: Embodied Question Answering in the Era of Foundation Models (CVPR 2024 · similarity 0.93)
- Multi-modal Situated Reasoning in 3D Scenes (NeurIPS 2024 · similarity 0.89)
- NavBench: Probing Multimodal Large Language Models for Embodied Navigation (NeurIPS 2025 · similarity 0.88)
- IndustryEQA: Pushing the Frontiers of Embodied Question Answering in Industrial Scenarios (NeurIPS 2025 · similarity 0.88)
- Towards Learning a Generalist Model for Embodied Navigation (CVPR 2024 · 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.
