[Call for Reviewers]

[Call for Papers]

Introduction

We are excited to announce the first International Workshop on Knowledge-Based Compositional Generalization (KBCG) held in conjunction with IJCAI 2023. The workshop aims to bring together researchers from academia and industry to discuss the latest advances and challenges in the area of knowledge-based compositional generalization in AI. Compositional generalization, the ability to understand and generate new combinations of previously learned concepts, is a fundamental problem in AI. It is particularly important in areas such as natural language understanding, reinforcement learning, and knowledge representation. However, achieving compositional generalization remains a challenging problem in AI.

The ability to make generalizations based on different actions or concepts, known as systematic compositionality, is crucial for human learning and understanding. It allows us to understand and learn new things, even when we have limited experience. In human daily life, we often encounter problems that require us to be able to make generalizations based on compositions of different actions or concepts. This ability is critical for how we are able to learn and understand new things, even when we have limited experience. However, while there have been significant advances in the language capabilities of machines, they still struggle with generalization and require large amounts of training data. These machine learning techniques, like neural networks, have been criticized in the past for lacking systematic compositionality. For a recent survey on compositional generalization, please refer to: https://arxiv.org/abs/2302.01067. Some topics of interest include but not limited to:

• Representational learning, Meta-learning, Transfer learning, Reinforcement learning, Self-supervised learning for compositional generalization
• Reasoning, commonsense and knowledge representation for compositional generalization
• Applications of compositional generalization such as foundation models, natural language processing, computer vision, and control systems
• Methods to learn compositional representations
• Combining knowledge from multiple sources and modalities
• Relational and causal machine Learning
• Using external knowledge for efficient machine learning
• Symbol grounding and Abstractions
• Neuro-inspired AI
• Neuro-symbolic AI
• Compositionality in reinforcement learning
• Benchmarks for compositional generalization
• Data visualization to study compositionality

We believe that incorporating knowledge can potentially solve many of the most pressing challenges tackling the compositional generalization in deep learning today. The primary goal of this workshop is to facilitate community building: we hope to bring researchers together to consolidate this line of research and foster collaboration in the community from AI, cognitive sciences and neuroscience to discuss novel approaches such as representation learning, meta-learning transfer learning, reinforcement learning, self-supervised learning, foundation models, knowledge graph, data visualization and neuro-symbolic AI.

This workshop will be a hybrid event held in conjunction with IJCAI 2023, taking place on Aug 21st, 2023 at Macao, S. A. R. and virtually. The session will cover invited talks, contributed talks, posters, and a panel discussion.

Key Dates

• Submission deadline: April 26th, 2023 May 10th, 2023 (11:59 pm AOE, FINAL EXTENSION)
• Acceptance notification: June 5st, 2023
• Camera ready for accepted submissions: June 15th, 2023 June 20th, 2023

Keynote and Invited Speakers

Organizing Committee

Technical Program Committee (TPC)

We would like to express our sincere gratitude to our technical program committee for generously volunteering their time and expertise to review submissions for our workshop. Their valuable contributions have been instrumental in ensuring the quality and rigor of the workshop’s program. We deeply appreciate their dedication and commitment to our workshop’s success:

Jinqi Luo, Hariram Veeramani, Tianwei Xing, Adam Dahlgren, Bo Xiong, Emanuele Sansone, Mats Leon Richter, Peter Belcak, Tim Klinger, Yuwei Sun, Bo Dong, Tarun Kumar, Walid S. Saba, Yajie Bao, Aayush Mudgal, Daiki Kimura, Iordanis Fostiropoulos, Mariya Hendriksen, Vinutha Magal Shreenath

Contact

For any questions, please contact us at kbcg.workshop@gmail.com.

Sponsors

• IBM Research
• Columbia University
• Mila - Quebec AI Institute