Research map / 3D vision / Gaussian splatting
E3VS-Bench: A Benchmark for Viewpoint-Dependent Active Perception in 3D Gaussian Splatting Scenes
Koya Sakamoto, Taiki Miyanishi, Daichi Azuma, Shuhei Kurita, Shu Morikuni, Naoya Chiba, Motoaki Kawanabe, Yusuke Iwasawa, Yutaka Matsuo
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 |
|---|---|---|---|
| give agent privileged goal viewpoint (VQA at Goal) instead of start viewpoint (ablation) | Gemini 2.5 Flash · E3VS-Bench static baselines | avg judge score: 2.14 -> 3.58 | Improved |
| use bird's-eye view image instead of start view (ablation) | Gemini 2.5 Flash · E3VS-Bench static baselines | avg judge score: 2.14 -> 1.48 | Hurt |
| use 2D visual search (SEAL/DyFo) at start instead of VQA at Start (ablation) | Gemini 2.5 Flash · E3VS-Bench static baselines | avg judge score: 2.14 -> 1.68 (SEAL), 1.86 (DyFo) | No clear effect |
| enable thinking (ablation) | GPT 5.1 · E3VS-Bench | avg judge score: 2.42 -> 2.70 | Improved |
| enable thinking (ablation) | Gemini 3.0 Flash · E3VS-Bench | avg judge score: 2.79 -> 2.79 | No clear effect |
| use multi-frame input (1, 3, 5 frames) instead of single frame (ablation) | Gemini 3.0 Flash, Qwen3-VL-8B · E3VS-Bench | avg judge score | No clear effect |
| initialize agent at goal viewpoint instead of unanswerable start (ablation) | Gemini 3.0 Flash · E3VS-Bench | avg judge score: 2.79 -> 3.41 | Improved |
| initialize agent at goal viewpoint (ablation) | Qwen3-VL-8B · E3VS-Bench | avg judge score: 2.46 -> 3.38 | Improved |
| use Gemini 3.0 Flash agent instead of Random Action | E3VS agent · E3VS-Bench | avg judge score: 2.28 -> 2.79 | Improved |
Closest papers by meaning
- EmbodiedScan: A Holistic Multi-Modal 3D Perception Suite Towards Embodied AI (CVPR 2024 · similarity 0.89)
- IR3D-Bench: Evaluating Vision-Language Model Scene Understanding as Agentic Inverse Rendering (NeurIPS 2025 · similarity 0.88)
- Abstract 3D Perception for Spatial Intelligence in Vision-Language Models (CVPR 2026 · similarity 0.88)
- 3D-Mem: 3D Scene Memory for Embodied Exploration and Reasoning (CVPR 2025 · similarity 0.88)
- Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering (NeurIPS 2026 · similarity 0.88)
- D3D-VLP: Dynamic 3D Vision-Language-Planning Model for Embodied Grounding and Navigation (CVPR 2026 · 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.
