Commonsense Learning

Commonsense learning aims to equip artificial intelligence systems with the everyday knowledge and reasoning abilities that humans possess effortlessly. Current research focuses on integrating commonsense knowledge into various AI models, including reinforcement learning agents and multi-modal systems (e.g., those combining vision and language), often employing techniques like contrastive learning, graph neural networks, and self-supervised learning to improve performance and generalization. These advancements are significant because they address a critical limitation of current AI, paving the way for more robust and adaptable systems capable of handling real-world complexities in applications like question answering and visual reasoning.

Papers

March 15, 2024

EXPLORER: Exploration-guided Reasoning for Textual Reinforcement Learning
Kinjal Basu, Keerthiram Murugesan, Subhajit Chaudhury, Murray Campbell, Kartik Talamadupula, Tim Klinger
Reinforcement Learning Complex Reasoning Text Based Game Text Based Reinforcement Learning Commonsense Learning

November 11, 2023

Visual Commonsense based Heterogeneous Graph Contrastive Learning
Zongzhao Li, Xiangyu Zhu, Xi Zhang, Zhaoxiang Zhang, Zhen Lei
Visual Question Answering Visual Reasoning Visual Commonsense Heterogeneous Graph Contrastive Commonsense Learning

November 9, 2023

Active Mining Sample Pair Semantics for Image-text Matching
Yongfeng Chena, Jin Liua, Zhijing Yang, Ruihan Chena, Junpeng Tan
Image Text Matching Text Image Pair Commonsense Learning

March 17, 2022

elBERto: Self-supervised Commonsense Learning for Question Answering
Xunlin Zhan, Yuan Li, Xiao Dong, Xiaodan Liang, Zhiting Hu, Lawrence Carin
Contrastive Learning Question Answering Commonsense Reasoning Commonsense Knowledge Commonsense Question Answering Commonsense Learning

January 28, 2022

Leveraging class abstraction for commonsense reinforcement learning via residual policy gradient methods
Niklas Höpner, Ilaria Tiddi, Herke van Hoof
Reinforcement Learning Knowledge Graph Strong Generalization Class Specific Residual Reinforcement Learning Commonsense Learning

Commonsense Learning

Papers

EXPLORER: Exploration-guided Reasoning for Textual Reinforcement Learning

Visual Commonsense based Heterogeneous Graph Contrastive Learning

Active Mining Sample Pair Semantics for Image-text Matching

elBERto: Self-supervised Commonsense Learning for Question Answering

Leveraging class abstraction for commonsense reinforcement learning via residual policy gradient methods