Contextualized Scene Imagination for Generative Commonsense Reasoning

Authors: PeiFeng Wang, Jonathan Zamora, Junfeng Liu, Filip Ilievski, Muhao Chen, Xiang Ren

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

Reproducibility Variable Result LLM Response
Research Type Experimental The experiments demonstrate the effectiveness of I&V in improving language models on both concept-to-sentence and concept-to-story generation tasks, while enabling the model to learn well from fewer task examples and generate SKGs that make common sense to human annotators 1.
Researcher Affiliation Academia Peifeng Wang1,3, Jonathan Zamora2 , Junfeng Liu1 , Filip Ilievski3, Muhao Chen1,3, Xiang Ren1,3 1Department of Computer Science, University of Southern California 2Department of Computer Science, University of California, San Diego 3Information Sciences Institute, University of Southern California
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code Yes Code and data are available at https://github.com/wangpf3/imagine-and-verbalize.
Open Datasets Yes We evaluate Concept2Sentence on the Common Gen (Lin et al., 2020) benchmark... We construct two benchmarks based on the Visual Story Telling (VIST) (Huang et al., 2016) and ROCStories (Mostafazadeh et al., 2016) datasets.
Dataset Splits Yes We search the optimal hyper-parameters based on the perplexity over the development set, where the learning rate is chosen from {2e 6, 1e 5, 3e 5, 1e 4}, the batch size is chosen from {8, 16, 32, 64, 128}.
Hardware Specification No The paper mentions using models like "T5-large" and "BART-large" but does not specify any particular hardware (e.g., GPU model, CPU type, memory) used for running the experiments.
Software Dependencies No The paper mentions software components like "T5-base", "T5-large", "BART-large", and "Adam optimizer" but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes We use the Adam optimizer with weight decay 1e 2. We search the optimal hyper-parameters based on the perplexity over the development set, where the learning rate is chosen from {2e 6, 1e 5, 3e 5, 1e 4}, the batch size is chosen from {8, 16, 32, 64, 128}.