When an Evaluation Rule Writes Training Labels: Measuring Human-Reference Forgiveness in NAVSIM
cs.LG, cs.CV, cs.RO
Submitted: 2026-09-27
Updated: 2026-09-27
Code: https://github.com/NVlabs/GTRS
Terminology
Sources
- NAVSIM: Data-Driven Non-Reactive Autonomous Vehicle Simulation and Benchmarking
- Designing and Interpreting Probes with Control Tasks
- FlowR2A: Learning Reward-to-Action Distribution for Multimodal Driving Planning
- Generalized Trajectory Scoring for End-to-end Multimodal Planning
- Driving on Memory
- Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks
- Confident Learning: Estimating Uncertainty in Dataset Labels
- When Shared Rollouts Fail in Defensive Driving Evaluation: A NAVSIM Score Basis Audit
- Hidden Biases of End-to-End Driving Datasets
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks