Research map / Large language models / Alignment and preferences
Steerable Adversarial Scenario Generation through Test-Time Preference Alignment
Tong Nie, Yuewen Mei, Yihong Tang, Junlin He, Jie Sun, Haotian Shi, Wei Ma, Jian Sun
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 map-compliance as hard feasibility precondition in HGPO instead of removing it (ablation) | SAGE HGPO · Map feasibility during fine-tuning | map feasibility rate: nearly 90% feasibility vs collapse without map constraint | Improved |
| Use hard-constraint decoupling rather than weighted map penalty (ablation) | SAGE HGPO · Map feasibility and reward | feasibility and reward | Improved |
| Use group-based pairs (HGPO) instead of single winner-loser DPO (ablation) | SAGE · Fine-tuning efficiency | sample efficiency, training loss stability | Improved |
| Mix expert weights rather than logits or trajectories at test time (ablation) | SAGE · Pareto front adversariality vs realism | realism at given adversariality | Improved |
| Increase w_adv from 0.0 to 1.0 (ablation) | SAGE · WOMD vs Replay policy | Attack success rate: 16.26% -> 76.15% | Improved |
| Increase w_adv from 0.0 to 1.0 (ablation) | SAGE · WOMD vs Replay policy | Behavioral realism penalty (lower better): 0.332 -> 1.429 | Hurt |
| Train ego with SAGE adversaries instead of CAT adversaries | TD3 ego agent · Closed-loop adversarial env | route completion: 0.58 -> 0.69 | Improved |
| Train ego with SAGE adversaries instead of no adversary | TD3 ego agent · Closed-loop adversarial env | collision rate: 0.44 -> 0.31 | Improved |
| Train ego with SAGE adversaries on out-of-distribution adversary test sets | TD3 ego agent · Cross-eval CAT and Rule scenarios | collision rate: 0.307 vs CAT-trained 0.298 on CAT scenarios | Mixed |
Closest papers by meaning
- DiffScene: Diffusion-Based Safety-Critical Scenario Generation for Autonomous Vehicles (AAAI 2025 · similarity 0.88)
- DriveDPO: Policy Learning via Safety DPO For End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.87)
- World Models as Adversaries: Multi-Agent Self-Play Fine-Tuning for Robust Motion Planning (NeurIPS 2026 · similarity 0.87)
- Direct Post-Training Preference Alignment for Multi-Agent Motion Generation Model Using Implicit Feedback from Pre-training Demonstrations (ICLR 2025 · similarity 0.86)
- Hybrid-Driving: An Autonomous Driving Decision Framework Integrating Large Language Models, Knowledge Graphs and Driving Rules (AAAI 2025 · similarity 0.86)
- Bidirectional Motion Transformer for Safety-Critical Traffic Scenario Generation (NeurIPS 2025 · similarity 0.86)
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.
