Research map / Point clouds / Point cloud classification
Exploiting Vision Language Model for Training-Free 3D Point Cloud OOD Detection via Graph Score Propagation
Tiankai Chen, Yushu Li, Adam Goodge, Fei Teng, Xulei Yang, Tianrui Li, Xun 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 |
|---|---|---|---|
| replace cosine distance with shortest-path manifold distance (ablation) | ULIP2 OOD scoring · ScanObjNN | AUROC: 73.2 -> 74.1 | Improved |
| replace shortest-path distance with score propagation (ablation) | ULIP2 OOD scoring · ScanObjNN | AUROC: 74.1 -> 76.4 | Improved |
| add prompt clustering to score propagation (ablation) | GSP · ScanObjNN | AUROC: 76.4 -> 77.2 | Improved |
| add prompt clustering to score propagation (ablation) | GSP · ScanObjNN | FPR95: 72.6 -> 73.6 | Hurt |
| add self-training negative prompts to score propagation (ablation) | GSP · ScanObjNN | AUROC: 76.4 -> 77.3 | Improved |
| add self-training negative prompts to score propagation (ablation) | GSP · ScanObjNN | FPR95: 72.6 -> 71.4 | Improved |
| combine prompt clustering and negative prompts (ablation) | GSP · ScanObjNN | AUROC: 76.4 -> 78.6 | Improved |
| add 1-shot labeled ID samples to zero-shot GSP | GSP · ScanObjNN | AUROC: 78.6 -> 82.9 | Improved |
| add 5-shot labeled ID samples instead of zero-shot | GSP · ScanObjNN | AUROC: 78.6 -> 86.6 | Improved |
| use all training samples in the graph (full-shot) | GSP · ScanObjNN | AUROC: 86.6 -> 91.8 | Improved |
| increase prompt clusters Nc beyond 6 (ablation) | GSP prompt clustering · ScanObjNN | AUROC | Hurt |
| raise pseudo-negative percentage m beyond 6 (ablation) | GSP self-training · ShapeNetCore | AUROC | Hurt |
Closest papers by meaning
- Exploiting Vision Language Model for Training-Free 3D Point Cloud Understanding via Improved Graph Score Propagation (TPAMI 2026 · similarity 0.94)
- Overcoming Shortcut Problem in VLM for Robust Out-of-Distribution Detection (CVPR 2025 · similarity 0.87)
- Training an Open-Vocabulary Monocular 3D Detection Model without 3D Data (NeurIPS 2024 · similarity 0.87)
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.86)
- OpenAD: Open-World Autonomous Driving Benchmark for 3D Object Detection (NeurIPS 2025 · similarity 0.86)
- Hierarchical Cross-Modal Alignment for Open-Vocabulary 3D Object Detection (AAAI 2025 · similarity 0.85)
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.
