Research map / Multi-agent systems and games / Human-AI and goal-conditioned
GameFormer: Game-theoretic Modeling and Learning of Transformer-based Interactive Prediction and Planning for Autonomous Driving
Zhiyu Huang, Haochen Liu, 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 level-k iterative decoding (0 to 4 levels) (ablation) | GameFormer planner · WOMD open-loop planning | collision rate: 0.0384 -> 0.0198 | Improved |
| increase decoding levels beyond 4 (ablation) | GameFormer planner · WOMD open-loop planning | collision rate: 0.0198 -> 0.0245 (level 5) | Hurt |
| feed previous-level agent future trajectories into the next decoding level (ablation) | GameFormer · WOMD open-loop planning | planning ADE: 0.9210 -> 0.8329 | Improved |
| add self-attention over agent futures (ablation) | GameFormer · WOMD open-loop planning | planning ADE: 0.8666 -> 0.8329 | Improved |
| add interaction loss (ablation) | GameFormer · WOMD open-loop planning | collision rate: 0.0417 -> 0.0198 | Improved |
| use independent decoder layers instead of one shared decoder (ablation) | GameFormer · WOMD open-loop planning | planning ADE: 0.9196 -> 0.8329 | Improved |
| supervise intermediate decoding level trajectories (ablation) | GameFormer · WOMD open-loop planning | planning ADE: 0.9584 -> 0.8329 | Improved |
| use 6 decoding layers rather than 8 for interaction prediction (ablation) | GameFormer · WOMD interaction prediction | minADE: 0.9839 -> 0.9133 | Improved |
| add cost-based refinement planner | GameFormer · WOMD closed-loop planning | success rate (%): 73.16 -> 94.50 | Improved |
| use marginal model with EM aggregation instead of joint model | GameFormer · WOMD interaction prediction | mAP: mAP 0.1376 -> 0.1923, minFDE 1.9373 -> 2.2146 | Mixed |
Closest papers by meaning
- BehaviorGPT: Smart Agent Simulation for Autonomous Driving with Next-Patch Prediction (NeurIPS 2024 · similarity 0.89)
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- Diffusion-Based Planning for Autonomous Driving with Flexible Guidance (ICLR 2025 · similarity 0.88)
- DriveTransformer: Unified Transformer for Scalable End-to-End Autonomous Driving (ICLR 2025 · similarity 0.87)
- UniMotion: A Unified Motion Framework for Simulation, Prediction and Planning (NeurIPS 2025 · similarity 0.87)
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 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.
