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..
DAMamba: Vision State Space Model with Dynamic Adaptive Scan
Authors: Tanzhe Li, Caoshuo Li, Jiayi Lyu, Hongjuan Pei, Baochang Zhang, Taisong Jin, Rongrong Ji
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4 Experiments 4.1 Image Classification on Image Net-1K 4.2 Object Detection and Instance Segmentation on COCO2017 4.3 Semantic Segmentation on ADE20K 4.4 Ablation Study Results: As shown in Table 2, we compare the proposed DAMamba with several state-of-the-art models. |
| Researcher Affiliation | Academia | 1Key Laboratory of Multimedia Trusted Perception and Effcient Computing, Ministry of Education of China, Xiamen University, China 2School of Informatics, Xiamen University, China 3Fujian Ocean Innovation Center, China 4School of Engineering Science, University of Chinese Academy of Sciences, China 5School of Artificial Intelligence, Beihang University, China EMAIL,EMAIL |
| Pseudocode | No | The paper does not include a figure, block, or section labeled "Pseudocode", "Algorithm", or "Algorithm X", nor does it present structured steps for a method or procedure formatted like code. |
| Open Source Code | Yes | Code is available at https://github.com/ltzovo/DAMamba. |
| Open Datasets | Yes | We conducted image classification experiments based on the Image Net-1K dataset [43], which consists of 1,281,167 training images and 50,000 validation images spanning 1,000 categories. The COCO 2017 dataset [33] consists of approximately 118K training images and 5K validation images and serves as a commonly used benchmark for object detection and instance segmentation tasks. We conducted semantic segmentation experiments using the ADE20K dataset and performed a comparative analysis of DAMamba and other models within the Uper Net [60] framework. |
| Dataset Splits | Yes | We conducted image classification experiments based on the Image Net-1K dataset [43], which consists of 1,281,167 training images and 50,000 validation images spanning 1,000 categories. The COCO 2017 dataset [33] consists of approximately 118K training images and 5K validation images |
| Hardware Specification | Yes | The experiments were conducted on 16 RTX 3090 GPUs. The inference throughput is measured on an NVIDIA RTX 3090 GPU with a batch size 128. |
| Software Dependencies | No | The paper mentions software libraries like PyTorch [40] and Timm [56] but does not specify their version numbers. |
| Experiment Setup | Yes | The optimizer used was Adam W [38], with a cosine decay learning rate schedule and linear warm-up over the first 20 epochs. The models were trained for 300 epochs on images with a resolution of 224x224. For data augmentation and regularization, we employed techniques such as Rand Augmentation [6], Repeated Augmentation [21], Mixup [65], Cut Mix [64], Random Erasing [68], weight decay, label smoothing [47], and stochastic depth [24]. For the object detection and instance segmentation tasks, we trained the model for 12 epochs (1x) and 36 epochs (3x). The model optimization employed the Adam W optimizer with a batch size of 16. To ensure a fair comparison, all models were trained for 160k iterations within the Uper Net framework. |