Abstract Reasoning with Distracting Features

Authors: Kecheng Zheng, Zheng-Jun Zha, Wei Wei

NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results demonstrated strong improvements over baseline algorithms and we are able to beat the state-of-the-art models by 18.7% in the RAVEN dataset and 13.3% in the PGM dataset.
Researcher Affiliation Collaboration Kecheng Zheng University of Science and Technology of China zkcys001@mail.ustc.edu.cn Zheng-jun Zha University of Science and Technology of China zhazj@ustc.edu.cn Wei Wei Google Research wewei@google.com
Pseudocode No The paper does not contain pseudocode or clearly labeled algorithm blocks.
Open Source Code Yes 1Full code are available at https://github.com/zkcys001/distracting_feature.
Open Datasets Yes PGM [34] dataset consists of 8 different subdatasets, which each subdataset contains 119, 552, 000 images and 1, 222, 000 questions. ... RAVEN [39] dataset consists of 1, 120, 000 images and 70, 000 RPM questions, equally distributed in 7 distinct figure configurations
Dataset Splits No The paper mentions evaluating the student model on a “held-out validation set” in section 3.1 and using a “neutral train/test split” for PGM in section 4.1, but it does not specify the size or percentage for a validation split.
Hardware Specification No The paper does not explicitly describe the hardware (e.g., specific GPU/CPU models, memory) used to run its experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow versions, or other libraries).
Experiment Setup No The paper describes the architecture of its models (LEN, teacher model using DDPG) but does not provide specific experimental setup details such as hyperparameter values (e.g., learning rate, batch size, number of epochs) or optimizer settings used during training.