List your company

Image Captioning on TextCaps (val)

163.7CIDErPaLI-X
-4.9703238.8190982.6085126.39791May 12, 2021Jan 21, 2022Oct 2, 2022Jun 13, 2023Feb 22, 2024Nov 2, 2024Jul 14, 2025

Evaluation results

MethodLinks
PaLI-XOCR pipeline input=true2023.05163.7—————————
PaLI-3 (Specialist)OCR pipeline input=true, Model Category=Specialist, Reference Citation=[8]2023.12161—————————
PaLIOCR pipeline input=true2023.05160.4—————————
PaLINumber of parameters=17B2022.09160—————————
SMoLA-PaLI-3FTOCR pipeline input=true, Model Category=Generalist, Backbone=PaLI-32023.12159.3—————————
PaLI-3 (Specialist)OCR pipeline input=false, Model Category=Specialist, Reference Citation=[8]2023.12158.8—————————
SMoLA-PaLI-3FTOCR pipeline input=false, Model Category=Generalist, Backbone=PaLI-32023.12156.7—————————
SMoLA-PaLI-XFTOCR pipeline input=true, Model Category=Generalist, Backbone=PaLI-X2023.12154.7—————————
CogVLMOCR pipeline input=false, Model Category=Generalist2023.12151.3—————————
PaLINumber of parameters=15B2022.09150.1—————————
GIT2Number of parameters=5.1B2022.09148.6—————————
GIT22022.05148.638.428.3—21.954.6————
PaLI-XOCR pipeline input=false2023.05147—————————
GIT2OCR pipeline input=false2023.05145—————————
SMoLA-PaLI-XFTOCR pipeline input=false, Model Category=Generalist, Backbone=PaLI-X2023.12144.6—————————
GITNumber of parameters=0.7B2022.09143.7—————————
GIT2022.05143.73727.6—21.154.1————
PaLINumber of parameters=3B2022.09143.6—————————
Human2021.05125.524.426.14718.8—————
TAPchallenge_winner=CVPR 2021 workshop challenge2022.0511928.124.4—17.749.3————
mPLUG-DocOwlOCR pipeline input=false, Model Category=Generalist2023.12111.9—————————
TAP2022.05109.225.823.8—17.147.9————
GITvariant=L2022.05106.330.624.6—18.650.3————
Anc.-Cap.2022.0595.524.722.5—15.947.1————
PixelM4C-CaptionerOCR Source=TextOCR, OCR Feature=MTS v3 LH (from 9)2021.0591.9924.121.9846.6515.08—————
PixelM4C-CaptionerOCR Source=TextOCR, OCR Feature=MTS v3 fc7 (from 8)2021.0591.4423.7621.8646.3815.14—————
M4C-CaptionerOCR Source=Rosetta (OCR-en), OCR Feature=Object detector fc72021.0589.623.32246.215.6—————
M4C-Cap.source=Sidorov et al. (2020)2022.0589.623.322—15.646.2————
PixelM4C-CaptionerOCR Source=MTS v3 (TextOCR-en), OCR Feature=MTS v3 LH2021.0586.8723.4121.4545.6914.75—————
PixelM4C-CaptionerOCR Source=MTS v3 (TextOCR), OCR Feature=MTS v3 LH2021.0585.3223.3321.345.7114.62—————
PixelM4C-CaptionerOCR Source=MTS v3 (TextOCR), OCR Feature=MTS v3 fc72021.0584.9923.0121.2745.6514.58—————
PixelM4C-CaptionerOCR Source=MTS v3 (TextOCR-en), OCR Feature=MTS v3 fc72021.0584.5423.1121.3745.5714.68—————
PixelM4C-CaptionerOCR Source=MTS v3 (COCO-Text+TT), OCR Feature=MTS v3 LH2021.0581.5523.0520.8845.4514.23—————
PixelM4C-CaptionerOCR Source=MTS v3 (COCO-Text+TT), OCR Feature=MTS v3 fc72021.0581.4423.0921.0845.5514.51—————
CAPSFUSIONScale=100M, Evaluation Protocol=Zero-shot2023.1066.2———15.2—————
GITvariant=B2022.0564.924.121.1—15.745.2————
CAPSFUSIONScale=50M, Evaluation Protocol=Zero-shot2023.1062.7———14.9—————
RawScale=100M, Evaluation Protocol=Zero-shot2023.1059.6———12.3—————
CAPSFUSIONScale=10M, Evaluation Protocol=Zero-shot2023.1056.3———13.9—————
Mixing (Raw & Syn.)Scale=50M, Evaluation Protocol=Zero-shot2023.1055.5———12.7—————
RawScale=50M, Evaluation Protocol=Zero-shot2023.1055.2———11.7—————
Mixing (Raw & Syn.)Scale=100M, Evaluation Protocol=Zero-shot2023.1055.1———12.3—————
Mixing (Raw & Syn.)Scale=10M, Evaluation Protocol=Zero-shot2023.1051.7———12.4—————
SyntheticScale=50M, Evaluation Protocol=Zero-shot2023.1046.7———13.2—————
SyntheticScale=100M, Evaluation Protocol=Zero-shot2023.1046.5———13.1—————
RawScale=10M, Evaluation Protocol=Zero-shot2023.1046.4———10.5—————
Mixing (Raw & LR)Scale=10M, Evaluation Protocol=Zero-shot2023.1044.6———10.3—————
AoANetsource=Sidorov et al. (2020)2022.0542.720.418.9—13.242.9————
SyntheticScale=10M, Evaluation Protocol=Zero-shot2023.1042.3———12.7—————
BUTDsource=Sidorov et al. (2020)2022.0541.920.117.8—11.742.9————
Language RewritesScale=10M, Evaluation Protocol=Zero-shot2023.1038.3———9.3—————
LoRA-MCLRank (r)=8, Alpha (α)=32, Decoding=BS, Beam=12025.071.6745233—25.559.759.982.86995.5
PaliGemma-3B (ft)Decoding=DBS (λ = 0.8), Beam=32025.071.66146.733.2—25.761.163.583.568.994.9
PaliGemma-3B (ft)Decoding=DBS (λ = 0.5), Beam=32025.071.65853.233—25.560.859.583.268.894.7
PaliGemma-3B (ft)Decoding=DBS (λ = 1.0), Beam=32025.071.65443.933.2—25.660.965.583.368.794.4
LoRA-MCLRank (r)=8, Alpha (α)=32, Decoding=BS, Beam=22025.071.6274933.3—25.860.461.882.468.793.2
LoRA-MoEDecoding=DBS (λ = 1.0), Beam=32025.071.62242.132.8—25.360.266.682.868.592.6
LoRA-MoEDecoding=DBS (λ = 0.8), Beam=32025.071.61644.132.7—25.460.365.482.868.592.4
LoRA-MoEDecoding=DBS (λ = 0.5), Beam=32025.071.6152.932.7—25.859.86082.268.492.3
LoRA-MoEDecoding=DBS (λ = 0.8), Beam=62025.071.60867.833—25.559.654.581.36892.2
LoRA-MoEDecoding=DBS (λ = 0.5), Beam=62025.071.60769.833.1—25.759.55381.468.192.3
LoRA-MoEDecoding=DBS (λ = 1.0), Beam=62025.071.6076733—25.459.655.181.367.992.1
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=DBS (λ = 1.0), Beam=32025.071.60142.532.7—25.260.166.582.768.691.5
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=DBS (λ = 0.8), Beam=32025.071.5943.732.7—25.160.165.582.468.690.9
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=DBS (λ = 0.5), Beam=32025.071.58953.132.6—25.559.659.782.168.591.2
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=DBS (λ = 0.8), Beam=32025.071.58953.132.6—25.559.659.782.168.591.2
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=DBS (λ = 1.0), Beam=32025.071.58641.632.6—2559.966.982.268.590.6
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=DBS (λ = 0.5), Beam=62025.071.58669.732.9—25.359.452.880.868.191.1
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=DBS (λ = 1.0), Beam=62025.071.58467.132.8—25.159.255.1816890.8
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=DBS (λ = 0.8), Beam=62025.071.58368.132.9—25.359.354.2816890.9
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=DBS (λ = 0.5), Beam=62025.071.5869.432.8—25.359.553.281.368.190.8
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=DBS (λ = 1.0), Beam=62025.071.57766.632.8—2559.255.381.26890.4
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=DBS (λ = 0.8), Beam=62025.071.57367.132.8—25.159.354.981.268.190.3
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=BS, Beam=62025.071.57278.432.7—24.858.545.679.667.290.3
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=DBS (λ = 0.5), Beam=32025.071.57152.932.5—25.359.66081.868.490.2
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=BS, Beam=62025.071.55778.632.6—24.658.345.779.567.189.5
LoRA-MoEDecoding=SR BS, Beam=12025.071.54159.731.1—24.357.655.680.967.888.3
LoRA-MLERank (r)=24, Alpha (α)=96, Decoding=BS, Beam=32025.071.51768.831.5—24.45850.980.26787.3
LoRA-MLERank (r)=8, Alpha (α)=32, Decoding=BS, Beam=32025.071.51783.331.7—24.157.342.178.86787.4
AdaLoRA2026.02—————38————
LoRArank=642026.02—————42.6————
MARS2026.02—————42.9————