KEST: Kernel Distance Based Efficient Self-Training for Improving Controllable Text Generation

Authors: Yuxi Feng, Xiaoyuan Yi, Laks V.S. Lakshmanan, Xing Xie

IJCAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on three controllable generation tasks demonstrate that KEST significantly improves control accuracy while maintaining comparable text fluency and generation diversity against several strong baselines.
Researcher Affiliation Collaboration 1The University of British Columbia, Vancouver, Canada 2Microsoft Research Asia, Beijing, China
Pseudocode Yes Algorithm 1 Training Process of KEST
Open Source Code Yes Code and appendices are available at https://github.com/peterfengyx/KEST.
Open Datasets Yes We evaluate the sentiment controllability on the IMDb movie review dataset [Maas et al., 2011]. We use the AGNews dataset [Zhang et al., 2015] to evaluate topic controllability. We use the Jigsaw Toxicity Dataset for training...
Dataset Splits Yes For IMDb, we sample 5% of the training samples as labeled data and directly take their provided unlabeled set. Since there is no separate unlabeled text in AGNews, we sample 3% of training samples as labeled data and use the others as unlabeled ones. For a fair comparison, we keep the ratio of labeled/pseudo/unlabeled text to 1:1:30.
Hardware Specification No The paper does not provide specific hardware details such as exact GPU/CPU models or types of processors used for experiments.
Software Dependencies No The paper mentions software like Uni LM, Adam W, RoBERTa-large, BERT-base, and GPT2-XL, but does not specify their version numbers for reproducibility.
Experiment Setup Yes We use Adam W [Loshchilov and Hutter, 2019] with learning rate = 5e-5, warm-up steps = one epoch, and batch size = 8 for optimization. The top-p (p = 0.9) sampling method is used for decoding in evaluation. We set λc = λag = λnag = 1.0 in Eq. (4) across all tasks.