AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

CorrectionPlanner: Self-Correction Planner with Reinforcement Learning in Autonomous Driving

Yihong Guo, Dongqiangzi Ye, Sijia Chen, Anqi Liu, Xianming Liu

ICML 2026 · Reinforcement learning › Visual and representation RL · approach: RL for combinatorial routing · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add self-correction at inference (ablation)CorrectionPlanner · Waymo collisioncollision rate: 2.20 -> 1.68Improved
add self-correction at inference (ablation)CorrectionPlanner · nuPlan collisioncollision rate: 2.45 -> 2.04Improved
add self-correction at inference to IL policy (ablation)IL · Waymo collisioncollision rate: 2.34 -> 2.31No clear effect
add RL training (ablation)IL · Waymo collisioncollision rate: 2.34 -> 2.21Improved
add correction-trace aggregation instead of conditioning on last unsafe token (ablation)Alternate Self-Correction · Waymo collisioncollision rate: 2.01 -> 1.68Improved
use self-correction instead of rejection sampling (RL) (ablation)Rej. Sampling RL · Waymo collisioncollision rate: 2.09 -> 1.68Improved
use self-correction instead of candidate selection (RL) (ablation)Candidate Sel. RL · Waymo collisioncollision rate: 2.14 -> 1.68Improved
lower collision critic threshold from 0.75 to 0.7 (CL=5) (ablation)CorrectionPlanner · Waymocollision rate / progression: 1.68 -> 1.47; 94.23 -> 77.50Mixed
raise correction length from 1 to 5 (threshold 0.75) (ablation)CorrectionPlanner · Waymocollision rate: 2.02 -> 1.68Improved
raise critic horizon k from 1 to 5 (ablation)collision critic · Waymo collisioncollision rate: 2.19 -> 1.68Improved
raise critic horizon k from 5 to 7 (ablation)collision critic · Waymocollision rate / progression: 1.68 -> 1.63; 94.23 -> 91.42Mixed
add self-correction to policy trained on nuPlan (zero-shot to Waymax) (ablation)CorrectionPlanner · Waymax zero-shotcollision rate: 2.93 -> 2.62Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy