Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
REDOUBT: Duo Safety Validation for Autonomous Vehicle Motion Planning
Authors: Shuguang Wang, Qian Zhou, Kui Wu, Dapeng Wu, Wei-Bin Lee, Jianping Wang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our extensive experiments demonstrate that both modules outperform existing approaches, under both open-loop and closed-loop evaluation settings. Our method demonstrates superior performance in both open-loop and closed-loop settings. |
| Researcher Affiliation | Collaboration | 1City University of Hong Kong, Hong Kong, China 2University of Victoria, B.C., Canada 3Information Security Center, Hon Hai Research Institute, Taipei, Taiwan |
| Pseudocode | No | The paper describes the methodology in narrative text and mathematical equations in Section 3. No explicit 'Pseudocode' or 'Algorithm' blocks are present. |
| Open Source Code | Yes | Our codes are available at: https://github.com/sgNicola/Redoubt. |
| Open Datasets | Yes | We conduct our evaluations using nu Plan [4], which is currently the most widely used and the only large-scale dataset for motion planning evaluation. |
| Dataset Splits | Yes | We partition the dataset based on scenario frequency, following a strategy similar to that used in Shifts [36]. Specifically, the top 50% most frequent scenario types which account for 95% of the total data are categorized as the In Distribution (In D) set, while the remaining, less common scenarios constitute the Out-of-Distribution (OOD) set. For more details, please refer to Appendix A.1 |
| Hardware Specification | Yes | Our experiments are conducted on a server with 8 NVIDIA RTX 5880 Ada GPUs and the Pytorch platform. |
| Software Dependencies | No | Our experiments are conducted on a server with 8 NVIDIA RTX 5880 Ada GPUs and the Pytorch platform. |
| Experiment Setup | Yes | The training dataset contains a total of 300,000 scenarios, including specific In D scenario types. The evaluation dataset consists of 19,685 scenarios, with fixed scenario tokens. For each scenario type (except for the "Unknown" scenario, which cannot be verified based on the scenario type), either 500 scenarios are included, or, if fewer than 500 scenarios are available, all the available scenarios are used. All planners are trained exclusively on the In D dataset and evaluated on the full dataset, which includes both In D and OOD scenarios. |