We hypothesize that relying heavily on simple co-occurrence statistics of the pre-training corpora is one of the main factors that cause factual errors. Our results reveal that LLMs are vulnerable to the co-occurrence bias, defined as preferring frequently co-occurred words over the correct answer. Consequently, LLMs struggle to recall facts whose subject and object rarely co-occur in the pre-training dataset although they are seen during finetuning.

Beyond Single Path Integrated Gradients
for Reliable Input Attribution via Randomized Path Sampling 
written by Giyoung Jeon, Haedong Jeong and Jaesik Choi is accepted at ICCV-2023.

Rarity Score: A New Metric to Evaluate the Uncommonness of Synthesized Images,
written by Jiyeon Han, Hwanil Choi, Yunjey Choi, Junho Kim, Jung-Woo Ha and Jaesik Choi is accepted at ICLR-23.

An Efficient Explorative Sampling Considering the Generative Boundaries of Deep Generative Neural Networks
Giyoung Jeon*, Haedong Jeong* and Jaesik Choi
In this paper, we present an explorative sampling algorithm to analyze generation mechanism of DGNNs. Our method efficiently obtains samples with identical attributes from a query image in a perspective of the trained model. We define generative boundaries which determine the activation of nodes in the internal layer and probe inside the model with this information.