Research map / Vision-language and multimodal / Robot action and navigation
Percept-WAM: Perception-Enhanced World-Awareness-Action Model for Robust End-to-End Autonomous Driving
Jianhua Han, Meng Tian, Jiangtong Zhu, Fan He, Huixin Zhang, Sitong Guo, Dechang Zhu, Hao Tang, Pei Xu, Yuze Guo, Minzhe Niu, Haojie Zhu and 7 more
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 IoU confidence trained on real model-prediction distribution (ablation) | Percept-WAM 2D detection · nuImages val | AP: 48.1 -> 49.6 | Improved |
| add IoU confidence trained on random-perturbed GT boxes (ablation) | Percept-WAM 2D detection · nuImages val | AP: 48.1 -> 46.9 | Hurt |
| add IoU confidence trained on uniformly sampled model predictions (ablation) | Percept-WAM 2D detection · nuImages val | AP: 48.1 -> 46.2 | Hurt |
| initialize World-BEV tokens from pretrained LiDAR encoder (ablation) | Percept-WAM camera baseline · nuScenes val BEV 3D detection | mAP: 25.0 -> 33.2 | Improved |
| add LiDAR and image data augmentation (ablation) | Percept-WAM BEV detection · nuScenes val | mAP: 33.2 -> 41.3 | Improved |
| increase BEV grid resolution from 20x20 to 40x40 (ablation) | Percept-WAM BEV detection · nuScenes val | mAP: 41.3 -> 50.4 | Improved |
| replace AR coordinate decoding with parallel MLP decoding (ablation) | Percept-WAM BEV detection · nuScenes val | mAP / NDS: mAP 50.4 -> 50.4; NDS 46.6 -> 43.7 | Mixed |
| replace AR trajectory decoding with query-based decoding (ablation) | Percept-WAM · nuScenes val | avg L2 / latency: L2 0.3970 -> 0.3822; latency 2700 -> 1174 ms | Improved |
| add streaming inference to query-based decoding (ablation) | Percept-WAM · nuScenes val | latency: 1174 -> 707 ms | Improved |
| add streaming inference to AR decoding (ablation) | Percept-WAM · nuScenes val | avg L2: 0.3970 -> 0.4058 | Hurt |
| joint training on AD and general datasets vs AD only (ablation) | Percept-WAM · nuImages 2D detection | mAP: 49.9 -> 49.6 | No clear effect |
| train on all PV tasks (2D + 3D) vs 2D AD only (ablation) | Percept-WAM · nuImages 2D detection | mAP: 46.7 -> 49.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SECOND — SECOND: Mitigating Perceptual Hallucination in Vision-Language Models via Selective and Contrastive Decoding (ICML 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SpaceDrive: Infusing Spatial Awareness into VLM-based Autonomous Driving (CVPR 2026 · similarity 0.89)
- DriveSpatial: A Benchmark for Spatiotemporal Intelligence in VLMs for Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- Explicit Geometric Chain-of-Thought for Vision-Language-Action in Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- World Knowledge-Enhanced Reasoning Using Instruction-Guided Interactor in Autonomous Driving (AAAI 2025 · similarity 0.88)
- Prune2Drive: A Plug-and-Play Framework for Accelerating Vision-Language Models in Autonomous Driving (CVPR 2026 · similarity 0.87)
- FutureSightDrive: Thinking Visually with Spatio-Temporal CoT for Autonomous Driving (NeurIPS 2025 · 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.
