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].
Vision-aware Multimodal Prompt Tuning for Uploadable Multi-source Few-shot Domain Adaptation
Authors: Kuanghong Liu, Jin Wang, Kangjian He, Dan Xu, Xuejie Zhang
AAAI 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments were conducted on Office Home and Domain Net datasets to demonstrate the effectiveness of the proposed VAMP in the UMFDA, which outperformed the previous prompt tuning methods. |
| Researcher Affiliation | Academia | Kuanghong Liu, Jin Wang*, Kangjian He*, Dan Xu, Xuejie Zhang School of Information Science and Engineering, Yunnan University, Kunming, China EMAIL, EMAIL |
| Pseudocode | Yes | The pseudo-training procedure of the VAMP is shown in Algorithm I, in Appendix A. |
| Open Source Code | Yes | Code https://github.com/lkh-meredith/VAMP-UMFDA |
| Open Datasets | Yes | VAMP is evaluated on two multi-source few-shot domain adaptation benchmarks, Office Home (Venkateswara et al. 2017), and Domain Net (Peng et al. 2019). |
| Dataset Splits | Yes | For example, in Office Home, each source domain has 25% of the data, and the target domain has 25% of the data. For few-shot settings, only 1%, 3%, or 6% of samples in the source domain are annotated. To ensure the fair comparison, the same splits for all methods are used as in MSFAN (Yue et al. 2021a). |
| Hardware Specification | Yes | All experiments are implemented on an NVIDIA RTX 3090 GPU with 24GB memory. |
| Software Dependencies | Yes | We implement our framework using PyTorch framework with version 1.10.1. |
| Experiment Setup | Yes | The batch size is set to 32. We use AdamW optimizer with a learning rate of 2e-5, and weight decay 0.0001. We train for 50 epochs on Office Home and 20 epochs on Domain Net. The threshold T is set to 0.7 for pseudo label selection. The temperature T is fixed to 0.01 based on previous work (Radford et al. 2021). |