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..
DrivingRecon: Large 4D Gaussian Reconstruction Model For Autonomous Driving
Authors: Hao LU, Tianshuo Xu, Wenzhao Zheng, Yunpeng Zhang, Wei Zhan, Dalong Du, Masayoshi TOMIZUKA, Kurt Keutzer, Yingcong Chen
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we evaluate the performance of Driving Recon in terms of reconstruction and novel view synthesis, as well as explore its potential applications. We also provide detailed information on the dataset setup, baseline methods, and implementation details. Table 1: Comparison to state-of-the-art methods on the Waymo Open Dataset. PSNR, SSIM, and Depth RMSE (D-RMSE) are reported. |
| Researcher Affiliation | Collaboration | Hao LU1,2,3 , Tianshuo XU1,2, Wenzhao ZHENG3, Yunpeng ZHANG4, Wei ZHAN3, Dalong DU4, Masayoshi Tomizuka3, Kurt Keutzer3, Yingcong CHEN1,2, 1The Hong Kong University of Science & Technology (Guangzhou), 2The Hong Kong University of Science & Technology, 3University of California, Berkeley, 4Phi Gent Robotics |
| Pseudocode | No | The paper describes the methodology, such as the Prune and Dilate Block, through textual descriptions and diagrams (Figure 2, Figure 3), but does not present any explicitly labeled pseudocode or algorithm blocks. |
| Open Source Code | No | Our code is available at Drive Recon. (This statement is ambiguous as it does not provide a direct link and 'Drive Recon' is the model name). The checklist justification for question 5 states: 'Our code will be made publicly available soon.' |
| Open Datasets | Yes | Following [48, 42, 65, 56], we use both the Waymo Open dataset [45] and the nu Scenes [1] to test the algorithm s performance. The checklist justification for question 5 states: 'Dataset is open.' |
| Dataset Splits | Yes | All training and testing validation are consistent with the official ones. |
| Hardware Specification | Yes | The model is trained on 24 NVIDIA A100 (80G) GPUs for 50000 iterations, all about 24 gpu days. |
| Software Dependencies | No | The paper mentions using the Adam W optimizer, but does not specify any software names with version numbers (e.g., Python, PyTorch, CUDA versions). |
| Experiment Setup | Yes | The model is trained on 24 NVIDIA A100 (80G) GPUs for 50000 iterations, all about 24 gpu days. A batch size of 2 for each GPU is used under bfloat16 precision, resulting in an effective batch size of 48. The Adam W optimizer is employed with a learning rate of 4 10 4 and a weight decay of 0.05. λre, λc, λr, λdr, λsr, λseg are set as 1.0, 0.1, 0.1, 0.1, 0.1, 0.1, respectively. |