KG-BART: Knowledge Graph-Augmented BART for Generative Commonsense Reasoning

Authors: Ye Liu, Yao Wan, Lifang He, Hao Peng, Philip S. Yu6418-6425

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on benchmark Common Gen dataset verify the effectiveness of our proposed approach by comparing with several strong pre-trained language generation models
Researcher Affiliation Academia 1University of Illinois at Chicago, Chicago, IL, USA 2Huazhong University of Science and Technology, Wuhan, China 3 Lehigh University, Bethlehem, PA, USA 4Beihang University, Beijing, China
Pseudocode No The paper describes methods using textual descriptions and diagrams (Figures 2, 3, 4), but does not contain a formal pseudocode or algorithm block.
Open Source Code Yes 1Our code is available at https://github.com/yeliu918/KG-BART
Open Datasets Yes Dataset Common Gen (Lin et al. 2020) is a constrained text generation task, which is to explicitly test the ability of machines on commonsense reasoning when generating a text. The dataset released in this task is constructed through a combination of crowdsourced and existing caption corpora, which consists of 77k commonsense descriptions over 35k unique concept sets.
Dataset Splits Yes Train Dev Test # Concept sets 32,651 993 1,497 # Sentences 67,389 4,018 6,042
Hardware Specification No The paper does not provide specific details about the hardware used for running experiments.
Software Dependencies No The paper mentions various models and techniques (e.g., GPT-2, Uni LM, T5, BART, RoBERTa, TransE, GloVe, CNN) but does not provide specific version numbers for software dependencies or libraries required for replication.
Experiment Setup No The paper describes the model architecture and pre-training objectives, but does not explicitly state specific hyperparameters (e.g., learning rate, batch size, number of epochs) or detailed training configurations used in their experiments.