Research map / Vision-language and multimodal / Multimodal LLMs
Less is More: A Simple yet Effective Object-Centric Prompting Strategy for Vision-Language Reasoning in Autonomous Driving
Saiqian Peng, Duanfeng Chu, Liping Lu, Bing Shi
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 |
|---|---|---|---|
| apply OCPS (box+ID overlay, ID output, lookup table) | DriveLM-agent · DriveLM-nuScenes | Accuracy: 0.6024 -> 0.6850 | Improved |
| apply OCPS | DriveLM-agent · DriveLM-nuScenes | Match: 32.29 -> 42.68 | Improved |
| apply OCPS | Qwen2.5-VL-7B · DriveLM-nuScenes | Match: 34.54 -> 54.56 | Improved |
| apply OCPS | InternVL-3.5-4B · DriveLM-nuScenes | Match: 37.88 -> 57.76 | Improved |
| apply OCPS | Qwen2.5-VL-7B · DriveLM-nuScenes | Accuracy: 0.7009 -> 0.8194 | Improved |
| apply OCPS | InternVL-3.5-4B · DriveLM-nuScenes | Chat: 71.65 -> 71.19 | No clear effect |
| add object-centric prompt generation without coordinate conversion (ablation) | DriveLM-agent · DriveLM-nuScenes | Match: 32.29 -> 28.27 | Hurt |
| add coordinate conversion on top of OCPG (ablation) | DriveLM-agent + OCPG · DriveLM-nuScenes | Match: 28.27 -> 42.68 | Improved |
| add object-centric prompt generation (IDs) to baseline (ablation) | DriveLM-agent · DriveLM-nuScenes | Accuracy: 0.6024 -> 0.6850 | Improved |
| use higher-resolution CLIP ViT-L/14-336 (ablation) | DriveLM-agent + OCPS · DriveLM-nuScenes | Match: 42.68 -> 46.12 | Improved |
| relax IoU threshold from 0.7 to 0.3 (ablation) | DriveLM-agent + OCPS · DriveLM-nuScenes | Match: 37.72 -> 42.68 | Improved |
| remove detection bounding boxes at evaluation (ablation) | DriveLM-agent + OCPS · DriveLM-nuScenes | Accuracy: 0.6850 -> 0.5745 | Hurt |
Closest papers by meaning
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.91)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.91)
- Prune2Drive: A Plug-and-Play Framework for Accelerating Vision-Language Models in Autonomous Driving (CVPR 2026 · similarity 0.90)
- LaST-VLA: Thinking in Latent Spatio-Temporal Space for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.90)
- Less Language, More Latents: Annotation-Efficient VLAs for Driving (NeurIPS 2026 · similarity 0.90)
- SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving (CVPR 2026 · similarity 0.90)
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.
