Research map / Image generation / Autoregressive and latent generation
Adversarial Generation and Collaborative Evolution of Safety-Critical Scenarios for Autonomous Vehicles
Jiangfan Liu, Yongkang Guo, Fangzhi Zhong, Tianyuan Zhang, Zonglei Jing, Siyuan Liang, Jiakai Wang, Mingchuan Zhang, Aishan Liu, Xianglong Liu
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 knowledge base D_l (license test Q&A) (ablation) | ScenGE meta-scenario generation · CARLA 8 scenarios | CR: 77.4% -> 82% (w/o D_l vs full) | Improved |
| add the knowledge base D_r (driving regulations) (ablation) | ScenGE · CARLA 8 scenarios | CR: 79.4% -> 82% | Improved |
| add the knowledge base D_c (pre-crash scenarios) (ablation) | ScenGE · CARLA 8 scenarios | CR: 80.5% -> 82% | Improved |
| swap qwq-32b for GPT-4o (ablation) | ScenGE LLM · CARLA 8 scenarios | CR: 82.0% (qwq-32b) vs 81.3% (GPT-4o) | No clear effect |
| perturb 4 collaborators rather than 2 (ablation) | Collaborator search · CARLA 8 scenarios | CR: 80.3% -> 82% | Improved |
| perturb 6 collaborators rather than 4 (ablation) | Collaborator search · CARLA 8 scenarios | CR: 82% -> 78.1% | Hurt |
| add temporal mask to attention (ablation) | Adversarial Collaborator Graph · CARLA 8 scenarios | CR: 79.3% -> 82% | Improved |
| add temporal decay to attention (ablation) | Adversarial Collaborator Graph · CARLA 8 scenarios | CR: 78.2% -> 82% | Improved |
| perturb 60% of segment instead of 50% or 70% (ablation) | Trajectory perturbation · CARLA 8 scenarios | CR: 80.2% (50%), 81.1% (70%) -> 82% (60%) | Improved |
| add L_ego term to loss (ablation) | Trajectory perturbation loss · CARLA 8 scenarios | CR: 78.6% -> 82% | Improved |
| add L_occ occlusion term to loss (ablation) | Trajectory perturbation loss · CARLA 8 scenarios | CR: 79.4% -> 82% | Improved |
| add L_smooth term to loss (ablation) | Trajectory perturbation loss · CARLA 8 scenarios | CR: 80.2% -> 82% | Improved |
Closest papers by meaning
- Safe-Sim: Safety-Critical Closed-Loop Traffic Simulation with Diffusion-Controllable Adversaries (ECCV 2024 · similarity 0.89)
- ChatScene: Knowledge-Enabled Safety-Critical Scenario Generation for Autonomous Vehicles (CVPR 2024 · similarity 0.88)
- DiffScene: Diffusion-Based Safety-Critical Scenario Generation for Autonomous Vehicles (AAAI 2025 · similarity 0.88)
- VILTA: A VLM-in-the-Loop Adversary for Enhancing Driving Policy Robustness (AAAI 2026 · similarity 0.88)
- World Models as Adversaries: Multi-Agent Self-Play Fine-Tuning for Robust Motion Planning (NeurIPS 2026 · similarity 0.86)
- Hybrid-Driving: An Autonomous Driving Decision Framework Integrating Large Language Models, Knowledge Graphs and Driving Rules (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.
