notesum.ai

Published at November 26

Relations, Negations, and Numbers: Looking for Logic in Generative Text-to-Image Models

cs.CV

cs.CL

cs.SC

Released Date: November 26, 2024

Authors: Colin Conwell¹, Rupert Tawiah-Quashie², Tomer Ullman

Aff.: ¹Harvard University; ²Hampshire College

Arxiv: http://arxiv.org/abs/2411.17066v1

Refer to caption

Model UID	Cohen’s Kappa
	Mean	LowerCI	UpperCI
retinanet_r50_fpn_ghm-1x_coco	-0.001	-0.001	-0.001
detr_r50_8xb2-150e_coco	0.012	0.004	0.022
gfl_r50_fpn_ms-2x_coco	0.018	0.012	0.023
dab-detr_r50_8xb2-50e_coco	0.019	0.013	0.028
ddq-detr-4scale_r50_8xb2-12e_coco	0.030	0.021	0.038
fovea_r50_fpn_4xb4-1x_coco	0.030	0.020	0.041
ddod_r50_fpn_1x_coco	0.032	0.019	0.044
faster-rcnn_regnetx-3.2GF_fpn_1x_coco	0.033	0.019	0.046
rtmdet_m_8xb32-300e_coco	0.033	0.019	0.048
deformable-detr_r50_16xb2-50e_coco	0.034	0.022	0.047
rtmdet_l_convnext_b_4xb32-100e_coco	0.037	0.023	0.051
rtmdet_l_swin_b_4xb32-100e_coco	0.038	0.023	0.053