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..
Adapting to Distribution Shift by Visual Domain Prompt Generation
Authors: Zhixiang Chi, Li Gu, Tao Zhong, Huan Liu, YUANHAO YU, Konstantinos N Plataniotis, Yang Wang
ICLR 2024 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments are conducted to validate the domain knowledge extraction. The proposed method outperforms previous work on 5 large-scale benchmarks including WILDS and Domain Net. |
| Researcher Affiliation | Academia | Zhixiang Chi1 , Li Gu2 , Tao Zhong1, Huan Liu3, Yuanhao Yu3, Konstantinos N Plataniotis1, Yang Wang2 1 University of Toronto, 2 Concordia University, 3 Mc Master University Q EMAIL |
| Pseudocode | Yes | Algorithm 1 Training scheme for VDPG |
| Open Source Code | No | Our source code will be available upon paper acceptance. |
| Open Datasets | Yes | We follow Meta-DMo E and MABN to evaluate VDPG on challenging real-world WILDS (Koh et al., 2021) benchmarks. |
| Dataset Splits | Yes | We follow official splits in source and target domains, and official metrics: accuracy, Macro F1, worse-case (WC) accuracy, Pearson correlation (r), and its worst-case. Specifically, for each episode, we first sample one domain Dn s p(s), and then sample two nonoverlapping support set (x S) and query set (x Q, y Q) (L4-5). |
| Hardware Specification | No | The paper specifies the model architectures (e.g., Vi T-B/16, Vi T-L/14) but does not provide any specific details about the hardware (GPUs, CPUs, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using CLIP and Vi T models, but does not provide specific version numbers for any software dependencies, such as deep learning frameworks or libraries. |
| Experiment Setup | Yes | We perform training using SGD with a batch size of 64 for 30 epochs. The initial learning rates are set to 3e 3 and 5e 4 with cosine decay for WILDS and Domain Net. The loss weights γ and λ are set to 0.1. |