Research map / Label-efficient and robust learning / Recognition and long-tail
Think Twice Before Driving: Towards Scalable Decoders for End-to-End Autonomous Driving
Xiaosong Jia, Penghao Wu, Li Chen, Jiangwei Xie, Conghui He, Junchi Yan, Hongyang Li
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 |
|---|---|---|---|
| Use supervised depth and segmentation predictions during image-to-BEV projection (ablation) | ThinkTwice encoder · Town05-Long | DS: 43.0$\pm$3.9 -> 57.8$\pm$1.5 | Improved |
| Add image depth and segmentation auxiliary tasks (ablation) | ThinkTwice encoder · Town05-Long | DS: 43.0$\pm$3.9 -> 45.0$\pm$2.2 | Improved |
| Use two input frames instead of one (ablation) | ThinkTwice encoder · Town05-Long | DS: 57.8$\pm$1.5 -> 59.3$\pm$1.4 | Improved |
| Use expert BEV feature distillation instead of BEV segmentation supervision (ablation) | ThinkTwice encoder · Town05-Long | DS: 50.5$\pm$2.3 -> 59.3$\pm$1.4 | Improved |
| Add one proposed decoder module (ablation) | ThinkTwice · Town05-Long | DS: 59.3$\pm$1.4 -> 61.6$\pm$1.4 | Improved |
| Stack five proposed decoder modules (ablation) | ThinkTwice · Town05-Long | DS: 59.3$\pm$1.4 -> 65.0$\pm$1.7 | Improved |
| Add the Look Module (ablation) | ThinkTwice decoder · Town05-Long | DS: 62.4$\pm$1.8 -> 65.0$\pm$1.7 | Improved |
| Add the Prediction Module (ablation) | ThinkTwice decoder · Town05-Long | DS: 61.7$\pm$1.9 -> 65.0$\pm$1.7 | Improved |
| Train the Prediction Module with teacher forcing (ablation) | ThinkTwice decoder · Town05-Long | DS: 62.0$\pm$2.2 -> 65.0$\pm$1.7 | Improved |
| Increase the TCP encoder from ResNet-34 to ResNet-101 (ablation) | TCP · Town05-Long | DS: 57.2$\pm$1.5 -> 38.0$\pm$4.5 | Hurt |
| Increase TCP encoder from ResNet-34 to ResNet-18 (ablation) | TCP · Town05-Long | DS: 57.2$\pm$1.5 -> 48.7$\pm$1.7 | Hurt |
| Increase the TCP MLP/GRU decoder depth and width (ablation) | ThinkTwice baseline · Town05-Long | DS: 59.3$\pm$1.4 -> 59.4$\pm$2.4 | No clear effect |
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.
- Transfuser — TransFuser: Imitation With Transformer-Based Sensor Fusion for Autonomous Driving (TPAMI 2022) · 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
- DriveAdapter: Breaking the Coupling Barrier of Perception and Planning in End-to-End Autonomous Driving (ICCV 2023 · similarity 0.90)
- DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving (ICLR 2025 · similarity 0.90)
- DualAD: Disentangling the Dynamic and Static World for End-to-End Driving (CVPR 2024 · similarity 0.89)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- FIVE-VLA: Fast and EffectIVE Closed-Loop Autonomous Driving with Recurrent Action Memory (NeurIPS 2026 · similarity 0.88)
- Future-Aware End-to-End Driving: Bidirectional Modeling of Trajectory Planning and Scene Evolution (NeurIPS 2025 · 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.
