Research map / Label-efficient and robust learning / Recognition and long-tail
ASSCG: Just-Right Gating over Chattering for Fast–Slow LLM Planning in Autonomous Driving
Sining Ang, Yuan Chen, Liu Haiyan, Xuanyao Mao, jason bao, Xuliang Xuliang, Bingchuan Sun, Yan Wang
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 ASSCG gate to AsyncDriver | AsyncDriver · nuPlan Hard20 | score: 65.00 -> 67.28 | Improved |
| add ASSCG gate to AsyncDriver | AsyncDriver · nuPlan Hard20 | latency (s/frame): 0.80 -> 0.32 | Improved |
| replace fixed 5-frame schedule with learned ASSCG at equal latency | AsyncDriver · nuPlan Hard20 | score: 64.27 -> 67.28 | Improved |
| add RL fine-tuning (GRPO) after SFT (ablation) | ASSCG · nuPlan Hard20 | score: 66.21 -> 67.28 | Improved |
| use GRPO instead of PPO (ablation) | ASSCG · nuPlan Hard20 | score: 66.76 -> 67.28 | Improved |
| use RWKV instead of Transformer backbone (ablation) | ASSCG · nuPlan Hard20 | last-frame gate latency (s): 0.06 -> 0.02 | Improved |
| use RWKV instead of Transformer backbone (ablation) | ASSCG · nuPlan Hard20 | score: 67.21 -> 67.28 | No clear effect |
| add Drop action (ablation) | ASSCG · nuPlan Hard20 | score: 66.71 -> 67.28 | Improved |
| raise query-cost weight lambda_call from 0.001 to 0.0020 (ablation) | ASSCG · nuPlan Hard20 | score: 67.28 -> 65.61 | Hurt |
| lower query-cost weight lambda_call from 0.001 to 0.0005 (ablation) | ASSCG · nuPlan Hard20 | score: 67.28 -> 67.25 | No clear effect |
| add confidence cue to gate input (ablation) | ASSCG · nuPlan Hard20 | score: 67.28 -> 67.33 | No clear effect |
| delay every Query by 3 frames (ablation) | ASSCG · nuPlan Hard20 | score: 67.28 -> 66.76 | Hurt |
Closest papers by meaning
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 2025 · similarity 0.85)
- AdaWM: Adaptive World Model based Planning for Autonomous Driving (ICLR 2025 · similarity 0.85)
- Drive-R1: Bridging Reasoning and Planning in VLMs for Autonomous Driving with Reinforcement Learning (AAAI 2026 · similarity 0.85)
- HCRMP: An LLM-Hinted Contextual Reinforcement Learning Framework for Autonomous Driving (NeurIPS 2025 · similarity 0.85)
- Hybrid-Driving: An Autonomous Driving Decision Framework Integrating Large Language Models, Knowledge Graphs and Driving Rules (AAAI 2025 · similarity 0.85)
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in 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.
