Image Captioning on COCO
154.9CIDErMethod [48]
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Method [48]SCST finetuning=true2023.05 | 154.9 | — | |
| SOTA2024.06 | 154.9 | — | |
| MiCo2024.06 | 152.4 | — | |
| PaLI-17B#Params=17B, Model Type=Specialist, Evaluation Protocol=Fine-tuned2023.11 | 149.1 | — | |
| Single Task SoTA2023.12 | 149.1 | — | |
| VASTSCST finetuning=true2023.05 | 149 | — | |
| BEIT-3CIDEr optimization=false2022.08 | 147.6 | — | |
| OFACIDEr optimization=false2022.08 | 145.3 | — | |
| PaliGemma2-28BResolution=448x448, Fine-tuning=per-task2025.12 | 145.2 | — | |
| PaliGemma2-3B + AuditDMResolution=448x448, Fine-tuning=per-task2025.12 | 145.1 | — | |
| GIT2#Params=5.1B, Model Type=Specialist, Evaluation Protocol=Fine-tuned2023.11 | 145 | — | |
| PaliGemma2-10BResolution=448x448, Fine-tuning=per-task2025.12 | 145 | — | |
| PaliGemma2-3BResolution=448x448, Fine-tuning=per-task2025.12 | 143.4 | — | |
| mPLUG-Owl2#Params=8.2B, Model Type=Generalist2023.11 | 137.3 | — | |
| mPLUG-Owl2Model Size=8.2B2023.12 | 137.3 | — | |
| PaLM-E-12B#Params=12B, Model Type=Generalist2023.11 | 135 | — | |
| Qwen-VL-Chat#Params=9.6B, Model Type=Generalist2023.11 | 131.9 | — | |
| Qwen-VLModel Size=9.6B2023.12 | 131.9 | — | |
| FinetunedEvaluation Protocol=Finetuned, Backbone=Vicuna 7B, LoRA rank=162026.03 | 130.5 | — | |
| UIO-2XLModel Size=3.2B2023.12 | 130.3 | — | |
| UIO-2LModel Size=1.1B2023.12 | 128.2 | — | |
| UIO-2XXLModel Size=6.8B2023.12 | 125.4 | — | |
| MM1-30BShot=16, Model Scale=30B2024.03 | 125.3 | — | |
| MM1-30BShot=8, Model Scale=30B2024.03 | 123.1 | — | |
| Unified-IOXL#Params=2.9B, Model Type=Generalist2023.11 | 122.3 | — | |
| LyricsBackbone=Vicuna-13B2023.12 | 121.1 | — | |
| CleanModel=LLaVa1.5-13B2026.02 | 119.2 | — | |
| Emu-14BModel size=14B, Category=Comprehension & Creation, Protocol=Zero-shot2023.09 | 117.7 | — | |
| Shikra#Params=7.2B, Model Type=Generalist2023.11 | 117.5 | — | |
| ShikraModel Size=7.2B2023.12 | 117.5 | — | |
| ShikraBackbone=Vicuna-13B2023.12 | 117.5 | — | |
| CaMML-13Bshots=32024.01 | 116.8 | — | |
| IDEFICS-80BShot=16, Training Data=Includes PMD (includes COCO), Model Scale=80B2024.03 | 116.6 | — | |
| MM1-7BShot=8, Model Scale=7B2024.03 | 116.3 | — | |
| CleanModel=LLaVa1.5-7B2026.02 | 115.5 | — | |
| DREAMLLM-7BModel size=7B, Category=Comprehension & Creation, Protocol=Zero-shot2023.09 | 115.4 | — | |
| HiFICLBackbone=Idefics1-9B2026.03 | 115.21 | — | |
| MimICBackbone=Idefics1-9B2026.03 | 114.89 | — | |
| GIT#Params=0.7B, Model Type=Specialist, Evaluation Protocol=Fine-tuned2023.11 | 114.8 | — | |
| MM1-3BShot=8, Model Scale=3B2024.03 | 114.6 | — | |
| IDEFICS-80BShot=8, Training Data=Includes PMD (includes COCO), Model Scale=80B2024.03 | 114.3 | — | |
| Flamingo-80Bshots=322024.01 | 113.8 | — | |
| CaMML-7Bshots=32024.01 | 111.4 | — | |
| Flamingo-80BShot=16, Model Scale=80B2024.03 | 110.5 | — | |
| Uni-Perceivershots=02024.01 | 109.8 | — | |
| Zero-ShotEvaluation Protocol=Zero-Shot2026.03 | 109.5 | — | |
| Flamingo-80BShot=8, Model Scale=80B2024.03 | 108.8 | — | |
| Flamingo-9Bshots=322024.01 | 106.3 | — | |
| 32-shot ICLBackbone=Idefics1-9B2026.03 | 105.89 | — | |
| DREAMLLM-7B*Model size=7B, Category=Comprehension & Creation, Protocol=Zero-shot, Training Data=SFT data by LLaVA-1.52023.09 | 103.7 | — | |
| Flamingo-80Bshots=42024.01 | 103.2 | — | |
| InstructBLIP#Params=8.2B, Model Type=Generalist2023.11 | 102.2 | — | |
| InstructBLIPModel Size=8.2B2023.12 | 102.2 | — | |
| KOSMOS-1shots=42024.01 | 101.7 | — | |
| AIMv2architecture=ViT-3B/14, #patches=576, resolution=336px2024.11 | 100.3 | — | |
| AIMv2architecture=ViT-H/14, #patches=576, resolution=336px2024.11 | 99.6 | — | |
| Flamingo-9BShot=8, Model Scale=9B2024.03 | 99 | — | |
| AIMv2architecture=ViT-1B/14, #patches=576, resolution=336px2024.11 | 98.4 | — | |
| IDEFICS-9BShot=8, Training Data=Includes PMD (includes COCO), Model Scale=9B2024.03 | 97 | — | |
| AIMv2architecture=ViT-L/14, #patches=576, resolution=336px2024.11 | 96.9 | — | |
| OpenAI CLIParchitecture=ViT-L/14, #patches=5762024.11 | 94.9 | — | |
| SigLIParchitecture=ViT-So400M/14, #patches=7522024.11 | 94.6 | — | |
| DINOv2architecture=ViT-g/14, #patches=30342024.11 | 93.4 | — | |
| Flamingo-9Bshots=42024.01 | 93.1 | — | |
| SigLIParchitecture=ViT-L/14, #patches=5762024.11 | 93 | — | |
| IDEFICS-80BBackbone=LLaMA-65B2023.12 | 91.8 | — | |
| IDEFICS-80BShot=0, Text-only demonstrations=True, Training Data=Includes PMD (includes COCO), Model Scale=80B2024.03 | 91.8 | — | |
| Flamingo-3BShot=8, Model Scale=3B2024.03 | 90.6 | — | |
| ReViLMshots=42024.01 | 90.5 | — | |
| ReViLMshots=82024.01 | 90.2 | — | |
| RA-CM3shots=22024.01 | 89.1 | — | |
| MoDESSkip x% Experts=50%, rho=0.482025.11 | 88.23 | — | |
| Kimi-VL-A3B-Instructk=6, Status=Default2025.11 | 86.7 | — | |
| MoDESSkip x% Experts=67%, rho=0.652025.11 | 85.37 | — | |
| Flamingo-80BBackbone=80B2023.12 | 84.3 | — | |
| Flamingo-80BShot=0, Text-only demonstrations=True, Model Scale=80B2024.03 | 84.3 | — | |
| MoDESSkip x% Experts=83%, rho=0.82025.11 | 81.33 | — | |
| CleanModel=OF-9B2026.02 | 79.7 | — | |
| Flamingo-9BModel size=9B, Category=Comprehension Only, Protocol=Zero-shot2023.09 | 79.4 | — | |
| Flamingo-9BBackbone=9B2023.12 | 79.4 | — | |
| Flamingo-9BShot=0, Text-only demonstrations=True, Model Scale=9B2024.03 | 79.4 | — | |
| ReViLMshots=22024.01 | 77.2 | — | |
| MM1-7BShot=0, Model Scale=7B2024.03 | 76.3 | — | |
| MIX.AttackModel=LLaVa1.5-13B, Perturbation (ε)=2/2552026.02 | 73.8 | — | |
| MM1-3BShot=0, Model Scale=3B2024.03 | 73.5 | — | |
| Flamingo-3BShot=0, Text-only demonstrations=True, Model Scale=3B2024.03 | 73 | — | |
| MM1-30BShot=0, Model Scale=30B2024.03 | 70.3 | — | |
| VT-AttackModel=LLaVa1.5-13B, Perturbation (ε)=2/2552026.02 | 68.6 | — | |
| MIX.AttackModel=LLaVa1.5-7B, Perturbation (ε)=2/2552026.02 | 67.5 | — | |
| VT-AttackModel=LLaVa1.5-7B, Perturbation (ε)=2/2552026.02 | 66.8 | — | |
| OF-9BModel size=9B, Category=Comprehension Only, Protocol=Zero-shot2023.09 | 65.5 | — | |
| Zero-shotBackbone=Idefics1-9B2026.03 | 63.06 | — | |
| CM3Leon-7B*Model size=7B, Category=Comprehension & Creation, Protocol=Zero-shot2023.09 | 61.6 | — | |
| ReViLMshots=02024.01 | 60.8 | — | |
| MIX.AttackModel=LLaVa1.5-13B, Perturbation (ε)=4/2552026.02 | 60.1 | — | |
| MIX.AttackModel=LLaVa1.5-7B, Perturbation (ε)=4/2552026.02 | 55.4 | — | |
| VT-AttackModel=OF-9B, Perturbation (ε)=2/2552026.02 | 49.7 | — | |
| AttackVLM-iiModel=LLaVa1.5-13B, Perturbation (ε)=2/2552026.02 | 46.6 | — | |
| IDEFICS-9BBackbone=LLaMA-7B2023.12 | 46 | — | |
| IDEFICS-9BShot=0, Text-only demonstrations=True, Training Data=Includes PMD (includes COCO), Model Scale=9B2024.03 | 46 | — |