List your company
ResearchFormerly WizWand

Hateful Meme Detection on Hateful Memes (val test)

51.05AccuracyQ-VLM
43.572445.513747.45549.3963Oct 10, 2024
Updated 8mo ago

Evaluation results

MethodLinks
Q-VLMBackbone=OpenFlamingo-3B, Shots=32, Bitwidth=8-bit2024.1051.05
Floating PointBackbone=OpenFlamingo-3B, Shots=32, Bitwidth=FP2024.1050.27
Floating PointBackbone=OpenFlamingo-3B, Shots=0, Bitwidth=FP2024.1050.23
Floating PointBackbone=OpenFlamingo-3B, Shots=4, Bitwidth=FP2024.1050.1
Q-LoRABackbone=OpenFlamingo-3B, Shots=32, Bitwidth=8-bit2024.1050.02
Q-VLMBackbone=OpenFlamingo-3B, Shots=4, Bitwidth=8-bit2024.1049.55
Q-VLMBackbone=OpenFlamingo-3B, Shots=0, Bitwidth=8-bit2024.1049.12
Q-LoRABackbone=OpenFlamingo-3B, Shots=4, Bitwidth=8-bit2024.1048.62
Q-VLMBackbone=OpenFlamingo-3B, Shots=32, Bitwidth=4-bit2024.1047.84
Q-LoRABackbone=OpenFlamingo-3B, Shots=0, Bitwidth=8-bit2024.1047.75
Q-LoRABackbone=OpenFlamingo-3B, Shots=32, Bitwidth=4-bit2024.1045.76
Q-VLMBackbone=OpenFlamingo-3B, Shots=4, Bitwidth=4-bit2024.1045.26
Q-LoRABackbone=OpenFlamingo-3B, Shots=4, Bitwidth=4-bit2024.1045.12
Q-VLMBackbone=OpenFlamingo-3B, Shots=0, Bitwidth=4-bit2024.1044.22
Q-LoRABackbone=OpenFlamingo-3B, Shots=0, Bitwidth=4-bit2024.1043.86