Self-Adversarial Learning with Comparative Discrimination for Text Generation
Authors: Wangchunshu Zhou, Tao Ge, Ke Xu, Furu Wei, Ming Zhou
ICLR 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on text generation benchmark datasets show that our proposed approach substantially improves both the quality and the diversity, and yields more stable performance compared to the previous GANs for text generation. We comprehensively evaluate the proposed self-adversarial learning paradigm in both synthetic data and real data on the text generation benchmark platform (Zhu et al., 2018). |
| Researcher Affiliation | Collaboration | Wangchunshu Zhou1 Tao Ge2 Ke Xu1 Furu Wei2 Ming Zhou2 1Beihang University, Beijing, China 2Microsoft Research Asia, Beijing, China |
| Pseudocode | Yes | Algorithm 1 Self-Adversarial Learning With Comparative Discriminator, Algorithm 2 Self-Adversarial Learning without self-play (i.e. CAL), Algorithm 3 Self-Adversarial Learning without comparative discriminator |
| Open Source Code | No | The paper does not contain an unambiguous statement that the authors are releasing the code for the work described in this paper, nor does it provide a direct link to a source-code repository. |
| Open Datasets | Yes | We evaluate our approach in both synthetic and real datasets based on Texygen (Zhu et al., 2018), which is a benchmark platform for evaluating adversarial text generation models. Table 1 presents the basic information of the datasets used for evaluation. Synthetic Image COCO EMNLP2017 WMT NEWS |
| Dataset Splits | No | The paper does not explicitly specify exact split percentages or absolute sample counts for training, validation, and test datasets, nor does it reference predefined splits with citations for reproducibility regarding data partitioning. |
| Hardware Specification | No | The paper does not provide specific hardware details such as exact GPU/CPU models, processor types, or memory amounts used for running its experiments. |
| Software Dependencies | No | The paper mentions software components like LSTM, Text CNN, and the Texygen platform, but does not provide specific version numbers for these or other relevant software dependencies such as programming languages or libraries. |
| Experiment Setup | Yes | We keep the most of the hyperparameters same with the Seq GAN except the hyperparameters introduced by our models (i.e., w(>), w(<), w( )) which are tuned based on the synthetic experiment and kept the same for the real data experiments. ... We choose batch size to be 64, dropout keep prob to be 0.75, l2 regularization to be 0.2. We pretrain all model for 120 epochs and fine-tune them until convergence. ... We tuned it based on the performance in synthetic experiment and find w0 : w2 = 1 : 0.1 to be a good choice for the initial weights1 and fixed w1 to be 0. |