Our new X account is live! Follow @wizwand_team for updates
Home
/
Benchmarks
Multimodal Question Answering on ScienceQA (SQA) (test)
Loading...
70.2
SQA Accuracy
Vanilla (LLaVA-NeXT-7B)
66.872
67.736
68.6
69.464
Jan 30, 2026
SQA Accuracy
CUDA Time
CUDA Time Reduction
FLOPs (T)
FLOPs Reduction
KV Cache (MB)
KV Cache Reduction
Updated 3d ago
Evaluation Results
Method
Method
Links
SQA Accuracy
CUDA Time
CUDA Time Reduction
FLOPs (T)
FLOPs Reduction
KV Cache (MB)
KV Cache Reduction
Vanilla (LLaVA-NeXT-7B)
#Tokens=2880
2026.01
70.2
-
-
9.6
-
1,512.1
-
VisionTrim
#Tokens=320
2026.01
69.6
-
61.4
0.8
91.7
101.8
93.3
VisionZip
#Tokens=320
2026.01
67.3
-
32.7
1.6
83.3
180.4
88.1
SparseVLM
#Tokens=320
2026.01
67
-
30.6
1.5
84.4
168
88.9
Feedback
Search any
task
Search any
task