Natural Language Queries on Ego4D NLQ (val) benchmark leaderboard · SOTA2 Research

Natural Language Queries on Ego4D NLQ (val)

21.97Recall@1 (IoU=0.3)OSGNet

Evaluation Results

Method	Links
OSGNet2025.05		21.97	44.61	15.2	32.96
NaQ2025.05		19.31	23.62	11.59	15.75
RGNet2025.05		18.66	36.37	11.72	22.43
RGNet2025.05		16.86	34.43	10.53	21.84
OSGNet2025.05		16.13	36.78	11.28	25.63
InternVideo2025.05		15.61	24.78	10.17	18.3
CONE2025.05		14.15	30.33	8.18	18.02
VSLNet2022.06		10.84	18.84	6.81	13.45
VSLNet2022.06		10.53	17.94	5.96	11.85
VSLNet2022.06		5.45	10.74	3.12	6.63
VSLNet2022.06		5.06	10.3	2.71	6.69
2D-TAN2022.06		5.04	12.89	2.02	5.88
VSLNet2022.06		3.95	8.72	2.01	4.62
Ours-full (+SW, VS, and Contra)2022.07		0.1079	0.1319	0.0674	0.0885
Ours-base + SW + VS2022.07		0.0996	0.1255	0.063	0.0834
Ours-base + SW2022.07		0.0906	0.1136	0.0568	0.0725
Ours-base2022.07		0.0769	0.1151	0.0483	0.078
Ours-base + VS2022.07		0.0738	0.1066	0.0431	0.0684
MS 2D-TAN2022.07		0.0705	0.1415	0.0475	0.0916
Ours-variant (self-attention)2022.07		0.0653	0.1102	0.0405	0.0759
VSLNet2022.07		0.0545	0.1074	0.0312	0.0663
2D-TAN2022.07		0.0504	0.1289	0.0202	0.0588
Moment DETR2022.07		0.0452	0.0803	0.0199	0.0333