Research map / Probabilistic and scientific modeling / Molecules and proteins
What Probing Reveals about Autonomous Driving: Better Predictions Lead to Better Planning
Hyeonchang Jeon, Kyungbeom Kim, Eugene Vinitsky, KyungJoong Kim
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Probes autonomous driving policies to test whether they actually predict surrounding vehicle motion and how the ego vehicle's actions affect it, finding that better predictions lead to better planning.
Problem
Closed-loop simulation reveals only resulting behavior, so it is unclear whether driving policies truly predict surrounding vehicles' motion or model how the ego vehicle's actions influence them, even when nominal performance looks strong.
Approach
Uses probing to inspect the internal representations of driving policies for prediction of other agents and ego-to-agent interaction, and relates the quality of these predictions to planning performance.
Key results
- Abstract indicates that better prediction of surrounding vehicle motion is associated with better planning performance.
Surprising findings
- Strong performance in nominal closed-loop scenarios can mask flawed reasoning and unsafe heuristics.
Closest papers by meaning
- Beyond Self-Play and Scale: A Behavior Benchmark for Generalization in Autonomous Driving (NeurIPS 2026 · similarity 0.89)
- Foresight in Motion: Reinforcing Trajectory Prediction with Reward Heuristics (ICCV 2025 · similarity 0.89)
- DIVER: Reinforced Diffusion Breaks Imitation Bottlenecks in End-to-End Autonomous Driving (TPAMI 2026 · similarity 0.89)
- Prioritizing Perception-Guided Self-Supervision: A New Paradigm for Causal Modeling in End-to-End Autonomous Driving (NeurIPS 2025 · similarity 0.88)
- DriveGPT: Scaling Autoregressive Behavior Models for Driving (ICML 2025 · similarity 0.88)
- Model-Based Policy Adaptation for Closed-Loop End-to-end Autonomous Driving (NeurIPS 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.
