List your company
ResearchFormerly WizWand

Text-to-Image Generation on MT Bench 90 prompts (test)

808Total WinsReFL
336.88459.19581.5703.81Apr 12, 2023
Updated 8mo ago

Evaluation results

MethodLinks
ReFLBackbone=Stable Diffusion v1.4, Scheduler=PNDM, Guidance Scale=7.52023.0480858.49
Dataset FilteringBackbone=Stable Diffusion v1.4, Scheduler=PNDM, Guidance Scale=7.52023.0473551.72
SD v1.4 (baseline)Backbone=Stable Diffusion v1.4, Scheduler=PNDM, Guidance Scale=7.52023.04718—
Reward WeightedBackbone=Stable Diffusion v1.4, Scheduler=PNDM, Guidance Scale=7.52023.0458543.33
RAFTBackbone=Stable Diffusion v1.4, Scheduler=PNDM, Guidance Scale=7.5, iter=12023.0457842.31
RAFTBackbone=Stable Diffusion v1.4, Scheduler=PNDM, Guidance Scale=7.5, iter=22023.0445233.02
RAFTBackbone=Stable Diffusion v1.4, Scheduler=PNDM, Guidance Scale=7.5, iter=32023.0435526.19