Research map / Reinforcement learning / Visual and representation RL
CIRL: Controllable Imitative Reinforcement Learning for Vision-based Self-driving
Xiaodan Liang, Tairui Wang, Luona Yang, Eric Xing
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 abnormal steer-angle reward (ablation) | CIRL · CARLA One turn (New town) | success rate (%): 65 -> 71 | Improved |
| add abnormal steer-angle reward (ablation) | CIRL · CARLA One turn (Training conditions) | success rate (%): 91 -> 97 | Improved |
| add abnormal steer-angle reward (ablation) | CIRL · CARLA One turn (New weather) | success rate (%): 96 -> 94 | No clear effect |
| add demonstration transitions to replay buffer (ablation) | CIRL · CARLA One turn (New town) | success rate (%): 71 -> 71 | No clear effect |
| add demonstration transitions to replay buffer (ablation) | CIRL · CARLA One turn (Training conditions) | success rate (%): 97 -> 96 | Hurt |
| run 0.5 million more simulation steps (ablation) | CIRL · CARLA One turn (New town) | success rate (%): 71 -> 68 | No clear effect |
| run 0.5 million more simulation steps (ablation) | CIRL · CARLA One turn (New weather) | success rate (%): 94 -> 98 | No clear effect |
| multiply reward scale by 10 (ablation) | CIRL · CARLA One turn (New town) | success rate (%): 71 -> 70 | No clear effect |
| divide reward scale by 10 (ablation) | CIRL · CARLA One turn (New town) | success rate (%): 71 -> 52 | Hurt |
| include speed reward (ablation) | CIRL · CARLA One turn (New town) | success rate (%): 20 -> 71 | Improved |
| include offroad and collision reward (ablation) | CIRL · CARLA One turn (New town) | success rate (%): 31 -> 71 | Improved |
| add imitative RL stage to imitation learning | IL · CARLA Navigation (New town) | success rate (%): 40 -> 53 | Improved |
Closest papers by meaning
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 2026 · similarity 0.88)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.88)
- Distill to Think, Foresee to Act: Cognitive-Physical Reinforcement Learning for Autonomous Driving (NeurIPS 2026 · similarity 0.87)
- AutoVLA: A Vision-Language-Action Model for End-to-End Autonomous Driving with Adaptive Reasoning and Reinforcement Fine-Tuning (NeurIPS 2025 · similarity 0.87)
- DriveAgent-R1: Advancing VLM-based Autonomous Driving with Active Perception and Hybrid Thinking (ICLR 2026 · 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.
