Self-Supervised Logic Induction for Explainable Fuzzy Temporal Commonsense Reasoning
Authors: Bibo Cai, Xiao Ding, Zhouhao Sun, Bing Qin, Ting Liu, Baojun wang, Lifeng Shang
AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experimental results on TIMEDIAL, a challenging dataset for temporal reasoning over dialog, show that our method, Logic Induction Enhanced Contextualized TEmporal Reasoning (LECTER), can yield great improvements over the traditional language model for temporal reasoning. |
| Researcher Affiliation | Collaboration | 1Research Center for Social Computing and Information Retrieval Harbin Institute of Technology, China 2Huawei Noah s Ark Lab |
| Pseudocode | No | No pseudocode or algorithm blocks were found in the paper. |
| Open Source Code | No | The paper mentions 'The implementation is based on Pytorch' but does not provide a link or explicit statement about releasing the source code for LECTER. It links to a dataset: 'https://github.com/qywu/Dialog Corpus'. |
| Open Datasets | Yes | We evaluate the performance of our proposed LECTER model on the challenge dataset TIMEDIAL (Qin et al. 2021). and We leverage other large-scale publicly available corpus containing over 700MB of text1 to construct our selfsupervised training dataset... 1https://github.com/qywu/Dialog Corpus |
| Dataset Splits | Yes | After preprocessing, we obtain 97k/24k instances for training/validation. |
| Hardware Specification | Yes | The implementation is based on Pytorch and trained on a Tesla V100 GPU with Adam optimizer with 10 epochs. |
| Software Dependencies | No | The implementation is based on Pytorch, but no specific version number for PyTorch or other software dependencies is provided. |
| Experiment Setup | Yes | During the training, the batch size is set to 32. The combination weight λ in Eq.7 is set to 1. We search the learning rate with grid search in lr {5e 6, 1e 5, 5e 5} for the baseline and LECTER. The implementation is based on Pytorch and trained on a Tesla V100 GPU with Adam optimizer with 10 epochs. |