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 [1].
Multi-Label Few-Shot Image Classification via Pairwise Feature Augmentation and Flexible Prompt Learning
Authors: Han Liu, Yuanyuan Wang, Xiaotong Zhang, Feng Zhang, Wei Wang, Fenglong Ma, Hong Yu
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental results demonstrate that our proposed method can push the performance to a higher level. ... Experiments Datasets We evaluate our method on two datasets, namely MS COCO (Lin et al. 2014) and PASCAL VOC (Everingham et al. 2015)... Baselines We compare our model with the following strong baselines... Ablation Study Ablation Study of Pairwise Feature Augmentation. |
| Researcher Affiliation | Academia | 1 Dalian University of Technology, Dalian, China 2 Peking University, Beijing, China 3 Shenzhen MSU-BIT University, Shenzhen, China 4 The Pennsylvania State University, Pennsylvania, USA |
| Pseudocode | No | The paper describes its methodology using descriptive text and mathematical equations in sections like "The Proposed Method" and "The Loss Function", but it does not include any clearly labeled pseudocode blocks or algorithms. |
| Open Source Code | No | The paper does not provide any explicit statements about releasing source code, nor does it provide a link to a code repository for the methodology described. |
| Open Datasets | Yes | We evaluate our method on two datasets, namely MS COCO (Lin et al. 2014) and PASCAL VOC (Everingham et al. 2015) |
| Dataset Splits | Yes | For the COCO dataset, which comprises 80 classes, we divide it into training/validation/test sets with a ratio of 52/12/16, respectively. Similarly, For the VOC dataset, which consists of 20 classes, we split it into training/validation/test sets with a ratio of 8/6/6, respectively. ... Following the settings of (Yan et al. 2022), during the construction of episodes, we set K = 1 and q = 4. |
| Hardware Specification | No | The paper mentions general computing resources in the acknowledgments: "Dalian Ascend AI Computing Center and Dalian Ascend AI Ecosystem Innovation Center for providing inclusive computing power and technical support." However, it does not specify any particular GPU models, CPU types, or other detailed hardware specifications used for running the experiments. |
| Software Dependencies | No | The paper mentions using "the visual and text encoders with frozen parameters from CLIP, where the visual encoder is ResNet50 (He et al. 2016) and the text encoder is Transformer (Vaswani et al. 2017)." While these are specific models/frameworks, no specific version numbers for these or any underlying software libraries (e.g., PyTorch, TensorFlow, Python) are provided. |
| Experiment Setup | Yes | In the prompt pool, we set the number of prompts M to 8, and the token number for each prompt T to 16. During the training process, we optimize the model parameters using the SGD optimizer with a learning rate of 0.02. ... α1 and α2 are both initialized to 1. ... uncertainty parameter σi initialized to 1. |