CLOSURE: Assessing Systematic Generalization of CLEVR Models

Abstract

The CLEVR dataset of natural-looking questions about 3D-rendered scenes has recently received much attention from the research community. A number of models have been proposed for this task, many of which achieved very high accuracies of around 97-99%. In this work, we study how systematic the generalization of such models is, that is to which extent they are capable of handling novel combinations of known linguistic constructs. To this end, we test models' understanding of referring expressions based on matching object properties (such as e.g. "another cube that is the same size as the brown cube") in novel contexts. Our experiments on the thereby constructed CLOSURE benchmark show that state-of-the-art models often do not exhibit systematicity after being trained on CLEVR. Surprisingly, we find that an explicitly compositional Neural Module Network model also generalizes badly on CLOSURE, even when it has access to the ground-truth programs at test time. We improve the NMN's systematic generalization by developing a novel Vector-NMN module architecture with vector-valued inputs and outputs. Lastly, we investigate how much few-shot transfer learning can help models that are pretrained on CLEVR to adapt to CLOSURE. Our few-shot learning experiments contrast the adaptation behavior of the models with intermediate discrete programs with that of the end-to-end continuous models.

Technical report

References in corpus (13)

• Sequence to Sequence Learning with Neural Networks
• Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
• VQA: Visual Question Answering
• Neural-Symbolic VQA: Disentangling Reasoning from Vision and Language Understanding
• Transparency by Design: Closing the Gap Between Performance and Interpretability in Visual Reasoning
• FiLM: Visual Reasoning with a General Conditioning Layer
• Compositional Attention Networks for Machine Reasoning
• Compositional generalization in a deep seq2seq model by separating syntax and semantics
• Measuring Compositional Generalization: A Comprehensive Method on Realistic Data
• C-VQA: A Compositional Split of the Visual Question Answering (VQA) v1.0 Dataset
• CLEVR: A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoning
• ShapeWorld - A new test methodology for multimodal language understanding
• Compositionality decomposed: how do neural networks generalise?

Cited by in corpus (10)

• Explainable Artificial Intelligence (XAI) 2.0: A Manifesto of Open Challenges and Interdisciplinary Research Directions
• Ground Truth Evaluation of Neural Network Explanations with CLEVR-XAI
• Compositional Generalization in Semantic Parsing: Pre-training vs. Specialized Architectures
• Learning to Recombine and Resample Data for Compositional Generalization
• Compositional Generalization via Neural-Symbolic Stack Machines
• Latent Compositional Representations Improve Systematic Generalization in Grounded Question Answering
• Visually Grounded Continual Learning of Compositional Phrases
• KANDINSKYPatterns -- An experimental exploration environment for Pattern Analysis and Machine Intelligence
• Iterated learning for emergent systematicity in VQA
• Neuro-Symbolic VQA: A review from the perspective of AGI desiderata