CSRR Workshop 2023

About

CSRR 2023 (TBA)

Common sense is the basic level of practical knowledge that is commonly shared among most people. Such knowledge includes but is not limited to social commonsense (“it's impolite to comment on people’s weight”), and physical commonsense (“snow is cold”). While humans use commonsense knowledge and reasoning abilities to seamlessly navigate everyday situations, endowing machines with such capabilities has remained an elusive goal of AI research for decades.

Recently, advances in large pre-trained LMs have shown that machines can directly learn large quantities of commonsense knowledge through self-supervised learning on raw data forms such as text and images. Additionally, their representations show promise at learning reasoning abilities when provided necessary supporting facts to reach a correct answer. However, despite these impressive exhibitions, these models still fall short of human-like understanding. While they achieve strong performance on benchmarks, they make inconsistent predictions, learn to exploit spurious patterns, generate socially stereotyped inferences, and fail to robustly apply learned knowledge to downstream applications.

Consequently, we organize this workshop to encourage discussion of current progress on building machines with commonsense knowledge and reasoning abilities. We aim to bring together researchers from different areas (e.g., NLP, computer vision, computational neuroscience, psychology) to communicate promising working directions in the area of commonsense reasoning.

Topics of interests include but not limited to

Methods: methods for commonsense reasoning tasks; methods that integrate commonsense knowledge bases and neural models; methods that improve the interpretability and explainability of neural models for reasoning and more.
Analysis: methods to probe commonsense knowledge from NLP models; methods to understand reasoning mechanisms of existing models; methods that identify limitations of existing methods for AI applications (including but not limited to NLP, CV and robotics) due to the lack of commonsense knowledge
Resources: acquiring commonsense knowledge (from text corpora, images, videos, pre-trained neural models, etc.); constructing and completing (semi-)structured CKBs; consolidating CKBs under unified schemas.
Benchmarks: designing challenging tasks and building datasets to evaluate models’ commonsense knowledge and reasoning abilities; designing new evaluation schemas and metrics for commonsense reasoning tasks, particularly for open-ended and generative tasks

The CSRR 2023 Workshop is still under preparation! Previously, we hosted our 1st CSRR at ACL 2022 and also a workshop at AKBC 2021, named CSKB.