Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Multi-image reasoning on BLINK (val)

52.6Accuracy

LLaVA-NeXT-Interleave-7B

35.54439.97244.448.828Sep 30, 2024
Updated 1mo ago

Evaluation Results

MethodLinks
2024.09
52.6
2024.09
52.1
2024.09
52.1
2024.09
51.1
2024.09
50.2
2024.09
49.8
2024.09
49.7
2024.09
49.1
2024.09
48.2
2024.09
46.8
2024.09
46.3
2024.09
45.2
2024.09
44.2
2024.09
43.7
2024.09
43.5
2024.09
41.5
2024.09
40.3
2024.09
40
2024.09
39.2
2024.09
37.1
2024.09
36.2