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..
One-Step Diffusion for Detail-Rich and Temporally Consistent Video Super-Resolution
Authors: Yujing Sun, Lingchen Sun, Shuaizheng Liu, Rongyuan Wu, Zhengqiang ZHANG, Lei Zhang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments show that DLo RAL achieves strong performance in both accuracy and speed. Code and models are available at https://github. com/yjsunnn/DLo RAL. ... Our DLo RAL model achieves state-of-the-art performance on Real-VSR benchmarks, producing visually realistic frame details and stable temporal consistency. ... 4 Experiment |
| Researcher Affiliation | Collaboration | Yujing Sun1,2, , Lingchen Sun1,2, , Shuaizheng Liu1,2, Rongyuan Wu1,2, Zhengqiang Zhang1,2, Lei Zhang1,2, 1The Hong Kong Polytechnic University 2OPPO Research Institute |
| Pseudocode | No | The paper describes the methodology in detailed paragraph form and mathematical equations, but it does not contain any explicitly labeled 'Pseudocode' or 'Algorithm' blocks or figures. |
| Open Source Code | Yes | Code and models are available at https://github. com/yjsunnn/DLo RAL. |
| Open Datasets | Yes | Training Datasets. ... REDS dataset [22], ... Pexels1, ... LSDIR [18] dataset. Testing Datasets. We evaluate our method on both synthetic and real-world datasets, including UDM10 [47], SPMCS [28], Real VSR [45], and Video LQ [8]. |
| Dataset Splits | No | The paper mentions the total number of sequences and frames for testing datasets (UDM10, SPMCS, Real VSR, Video LQ) and states that 44,162 high-quality frames from REDS were selected for consistency stage training, and LSDIR for enhancement stage training. However, it does not provide explicit training/validation/test splits (e.g., percentages or counts) for the training data or for how the test sets are used in terms of splits within the experimental setup. |
| Hardware Specification | Yes | All models are trained using the Py Torch framework on 4 NVIDIA A100 GPUs. ... Inference time is measured on an A100 GPU (512 × 512 input with 50 frames for 4× VSR). ... measured on a single NVIDIA A100 80G GPU. |
| Software Dependencies | No | The paper mentions using 'Py Torch framework' for training and 'Stable Diffusion V2.1' as the backbone, but it does not provide specific version numbers for PyTorch or any other software libraries/dependencies. It also refers to Spy Net [24], which is an architecture/model rather than a specific software dependency with a version. |
| Experiment Setup | Yes | Training is carried out with a batch size of 16, a sequence length of 3, and a video resolution of 512 × 512. All models are trained using the Py Torch framework on 4 NVIDIA A100 GPUs. We use Adam optimizer with an initial learning rate of 5 × 10−5. |