Research map / Multi-agent systems and games / Human-AI and goal-conditioned
DecompGAIL: Learning Realistic Traffic Behaviors with Decomposed Multi-Agent Generative Adversarial Imitation Learning
Ke Guo, Haochen Liu, XIAOJUN WU, Chen Lv
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 DecompGAIL fine-tuning after BC pretraining (ablation) | SMART-tiny · WOSAC 2% validation split | realism meta-metric: 0.7836 -> 0.7889 | Improved |
| add DecompGAIL fine-tuning after BC pretraining (ablation) | SMART-tiny · WOSAC 2% validation split | collision likelihood: 0.9667 -> 0.9837 | Improved |
| add scene realism term (ablation) | DecompGAIL · WOSAC 2% validation split | map-based score: 0.8795 -> 0.8948 | Improved |
| add interaction realism term (ablation) | DecompGAIL · WOSAC 2% validation split | interactive score: 0.8132 -> 0.8283 | Improved |
| add interaction realism term (ablation) | DecompGAIL · WOSAC 2% validation split | collision likelihood: 0.9573 -> 0.9837 | Improved |
| use distance-weighted instead of uniform mean interaction realism (ablation) | DecompGAIL · WOSAC 2% validation split | realism meta-metric: 0.7819 -> 0.7889 | Improved |
| add neighborhood social reward (ablation) | DecompGAIL · WOSAC 2% validation split | realism meta-metric: 0.7871 -> 0.7889 | Improved |
| use distance-weighted instead of uniform mean neighborhood reward (ablation) | DecompGAIL · WOSAC 2% validation split | realism meta-metric: 0.7882 -> 0.7889 | Improved |
| replace standard local discriminator with decomposed discriminator (ablation) | PS-GAIL · WOSAC 2% validation split (gradient penalty) | realism meta-metric: 0.7674 -> 0.7889 | Improved |
| add R1 gradient penalty (ablation) | PS-GAIL · WOSAC 2% validation split (gradient penalty) | realism meta-metric: 0.7674 -> 0.7776 | Improved |
| add WGAN-GP gradient penalty (ablation) | DecompGAIL · WOSAC 2% validation split (gradient penalty) | realism meta-metric: 0.7889 -> 0.7865 | Hurt |
| add R2 gradient penalty (ablation) | DecompGAIL · WOSAC 2% validation split (gradient penalty) | realism meta-metric: 0.7889 -> 0.7840 | Hurt |
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.
- SMART-tiny — SMART: Simultaneous Multi-Agent Recurrent Trajectory Prediction (ECCV 2020) · better on 2 of 3 shared results
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
- LASIL: Learner-Aware Supervised Imitation Learning For Long-term Microscopic Traffic Simulation (CVPR 2024 · similarity 0.88)
- Flow Matching-Based Autonomous Driving Planning with Advanced Interactive Behavior Modeling (NeurIPS 2025 · similarity 0.86)
- Closed-Loop Supervised Fine-Tuning of Tokenized Traffic Models (CVPR 2025 · similarity 0.86)
- BehaviorGPT: Smart Agent Simulation for Autonomous Driving with Next-Patch Prediction (NeurIPS 2024 · similarity 0.85)
- Trajeglish: Traffic Modeling as Next-Token Prediction (ICLR 2024 · similarity 0.85)
- MAPLE: Latent Multi-Agent Play for End-to-End Autonomous Driving (NeurIPS 2026 · 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.
