Debiased Active Learning with Variational Gradient Rectifier
Authors: Weiguo Chen, Changjian Wang, Shijun Li, Kele Xu, Yanru Bai, Wei Chen, Shanshan Li
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments demonstrate that our proposed methodology effectively counteracts bias phenomena in a majority of active learning scenarios. |
| Researcher Affiliation | Academia | 1College of Computer Science and Technology, National University of Defense Technology 2Academy of Medical Engineering and Translational Medicine, Tianjin University EMAIL, EMAIL, yr56 EMAIL |
| Pseudocode | No | The paper describes its methods using mathematical equations and textual explanations, but it does not contain any explicitly labeled "Pseudocode" or "Algorithm" blocks. |
| Open Source Code | No | The paper does not contain any explicit statement about releasing source code or a link to a code repository. |
| Open Datasets | Yes | We conduct bias rectification validations on six text datasets: SST-2 (Socher et al. 2013), AGNews (Zhang, Zhao, and Le Cun 2015), Pub Med (Dernoncourt and Lee 2017), TREC (Li and Roth 2002), SST-5 (Socher et al. 2013) and Newsgroup (Lang 1995). We also conduct experiments on two image datasets: Cifar-10 and Cifar-100 (Krizhevsky and Hinton 2009). For OOD generalization tasks, we employ the same datasets as in (Deng et al. 2023): SA (Kaushik, Hovy, and Lipton 2020), NLI (Kaushik, Hovy, and Lipton 2020), and ANLI (Houlsby et al. 2011). |
| Dataset Splits | No | The paper describes an iterative active learning process where "the most informative unlabeled data is selected from the unlabeled data pool and sent to oracle for labeling, updating the labeled data pool". While it mentions "consistent initial labeled samples" and reports accuracies at specific "checkpoints", it does not provide specific initial training/validation/test split percentages or sample counts for the datasets used to set up the experiments. |
| Hardware Specification | No | The paper does not provide any specific hardware details such as GPU models, CPU types, or memory amounts used for running the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers, such as programming languages, libraries, or frameworks. |
| Experiment Setup | No | The paper mentions several hyperparameters such as "EMA decay rate m", "inconsistency relaxation coefficient ϵu", "fluctuation relaxation coefficient ϵf", and "hyperparameters λ and δ". However, it explicitly states "Implementation details are provided in the Appendix.", indicating that concrete values for these or other training configurations (e.g., learning rate, batch size) are not present in the main text. |