List your company

Image Captioning on MS-COCO Karpathy 2014 (test)

41.3BLEU-4GIT
23.6228.2132.837.39Feb 19, 2015Jul 5, 2016Nov 20, 2017Apr 7, 2019Aug 22, 2020Jan 7, 2022May 25, 2023

Evaluation results

MethodLinks
GITPretrain Data=4M2023.0541.3139.1—30.4—24.3
ViTCapPretrain Data=4M2023.0541.2138.1—30.1—24.1
HAAVPretrain Data=None2023.0541141.5—30.2—23.9
VinVLPretrain Data=8.8M2023.0540.9140.4—30.9—25.1
OscarPretrain Data=6.5M2023.0540.5137.6—29.7—22.8
X-VLMPretrain Data=16M2023.0540.4139.3————
CLIP-ViLPretrain Data=None2023.0540.2134.2—29.7—23.8
DLCTPretrain Data=None2023.0539.8133.8—29.5—23
Xmodal-CtxPretrain Data=None2023.0539.7135.9—30—23.7
VLPPretrain Data=3M2023.0539.5129.3—29.3—23.2
X-LANPretrain Data=None2023.0539.5132—29.5—23.4
M²Pretrain Data=None2023.0539.1131.2—29.1—22.6
SimVLMbasePretrain Data=1.8B2023.0539134.8—32.9—24
AoANetPretrain Data=None2023.0538.9129.8—29.2—22.4
Up-DownPretrain Data=None2023.0536.3120.1—27.7—21.4
SCSTPretrain Data=None2023.0534.2114—26.7——
3 LBL ensembleOptimization method=DNGO, Ensemble Size=32015.0226.7—————
2 LBL ensembleOptimization method=DNGO, Ensemble Size=22015.0225.9—————
Single LBLOptimization method=DNGO, Ensemble Size=12015.0225.1—————
Hard-Attention LSTM2015.0225—————
10 LSTM ensembleEnsemble Size=102015.0224.4—————
Human Expert LBLHyperparameter Tuning=Human Expert2015.0224.3—————
Regularized LSTM2015.0224.3—————
Soft-Attention LSTM2015.0224.3—————
AnyGPTParameters=8B, zero-shot=true2024.02—107.5————
DreamLLMParameters=8B, zero-shot=true2024.02—115.4————
EmuParameters=14B, zero-shot=true2024.02—112.4————
FlamingoParameters=9B, zero-shot=true2024.02—79.4————
FlamingoParameters=80B, zero-shot=true2024.02—84.3————
InstructBLIPParameters=14B, zero-shot=false, trained on training samples=true2024.02—102.2————
SEED-LLaMAParameters=8B, zero-shot=false, trained on training samples=true2024.02—123.6————