Research map / Detection and segmentation / 3D detection and driving
Agent-OBJ: Prompt-Driven 3D Adversaries for Multi-Modal Perception
Bing Li, Sean Du, Xuhong Ren, Luqi Gong, Wee Peng Tay, Qing Guo
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 the joint objective over both detectors (ablation) | Agent-OBJ · nuScenes | ASR_joint: 40.8 -> 64.0 | Improved |
| add multi-view personalization (ablation) | Agent-OBJ · nuScenes | ASR_joint: 38.4 -> 64.0 | Improved |
| add multi-view stability regularizer L_mv (ablation) | Agent-OBJ · nuScenes | ASR_joint: 30.5 -> 64.0 | Improved |
| add naturalness regularizer L_nat (ablation) | Agent-OBJ · nuScenes | ASR_joint: 80.4 -> 64.0 | Hurt |
| add naturalness regularizer L_nat (ablation) | Agent-OBJ · nuScenes | FeatureDist: 0.36 -> 0.12 | Improved |
| use LLM agent guidance instead of random prompt edits | Agent-OBJ · nuScenes | ASR_joint: 13.9 -> 64.0 | Improved |
| optimize with joint feedback instead of 2D-only | Agent-OBJ · nuScenes | ASR_joint: 20.3 -> 64.0 | Improved |
| optimize with joint feedback instead of fusion-only | Agent-OBJ · nuScenes | ASR_joint: 13.7 -> 64.0 | Improved |
| use agent instead of greedy search | Agent-OBJ · nuScenes | ASR_joint@200: 31.5 -> 37.0 | Improved |
| use agent instead of greedy search | Agent-OBJ · nuScenes | Queries@80%: 648 -> 584 | Improved |
| swap 2D detector YOLOv10x to DINO without re-optimization | Agent-OBJ · nuScenes transfer | ASR_joint: 64.0 -> 53.7 | Hurt |
| swap fusion detector MambaFusion to BEVFusion | Agent-OBJ · nuScenes transfer | ASR_joint: 64.0 -> 66.2 | Improved |
Closest papers by meaning
- OBJVanish: Prompt-Driven Generation of Physically Realizable 3D LiDAR-Invisible Objects (ICML 2026 · similarity 0.88)
- Agent3D-Zero: An Agent for Zero-shot 3D Understanding (ECCV 2024 · similarity 0.86)
- E3VS-Bench: A Benchmark for Viewpoint-Dependent Active Perception in 3D Gaussian Splatting Scenes (ECCV 2026 · similarity 0.85)
- Achieving Speed-Accuracy Balance in Vision-based 3D Occupancy Prediction via Geometric-Semantic Disentanglement (AAAI 2025 · similarity 0.84)
- CorrBEV: Multi-View 3D Object Detection by Correlation Learning with Multi-modal Prototypes (CVPR 2025 · similarity 0.84)
- OpenAD: Open-World Autonomous Driving Benchmark for 3D Object Detection (NeurIPS 2025 · similarity 0.84)
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.
