Grape: Grammar-Preserving Rule Embedding

Authors: Qihao Zhu, Zeyu Sun, Wenjie Zhang, Yingfei Xiong, Lu Zhang

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We conducted experiments on six widely-used benchmarks containing four context-free languages. The results show that our approach improves the accuracy of the base model by 0.8 to 6.4 percentage points.
Researcher Affiliation Academia Qihao Zhu , Zeyu Sun , Wenjie Zhang , Yingfei Xiong , Lu Zhang Key Laboratory of High Confidence Software Technologies, Ministry of Education(Peking University); School of Computer Science, Peking University, 100871, P. R. China {zhuqh, szy , zhang wen jie,xiongyf,zhanglucs}@pku.edu.cn
Pseudocode No The paper does not contain any explicit pseudocode or algorithm blocks.
Open Source Code Yes The code is available at https://github.com/pkuzqh/Grape
Open Datasets Yes We evaluated our approach on six benchmarks, including the Hearth Stone benchmark [Ling et al., 2016], two semantic parsing benchmarks [Dong and Lapata, 2016], the Django benchmark [Yin and Neubig, 2017], the Concode benchmark [Iyer et al., 2018] and the Str Reg benchmark [Ye et al., 2020]. For this task, we adopted the widely used Java benchmark [Alon et al., 2019; Alon et al., 2018], Java-small, which contains 11 relatively large Java projects.
Dataset Splits Yes The statistics of these datasets are shown in Table 1. # Train, # Dev, # Test. This dataset contains 691,607 examples in the training set, 23,844 examples in the validation set and 57,088 examples in the test set.
Hardware Specification Yes It takes 34.78s for an epoch on a single Nvidia Titan RTX with Grape on average, whereas 30.74s without Grape.
Software Dependencies No The paper mentions using Adam optimizer and Tree Gen as the base model, and refers to official parsers for Java and Python (https://docs.python.org/3/library/ast.html and https://github.com/c2nes/javalang), but does not specify version numbers for general software dependencies or these parsers.
Experiment Setup Yes For the hyperparameters of our model, we set the number of iterations N = 9. The hidden sizes were all set to 256. We applied dropout after each iteration of the GNN layer, where the drop rate is 0.15. The model was optimized by Adam with learning rate lr = 0.0001.