Image Captioning on NoCaps
127CIDErPALI-17B
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PALI-17Bmode=specialist2023.08 | 127 | — | |
| LyricsBackbone=Vicuna-13B, Evaluation protocol=0-shot2023.12 | 126.8 | — | |
| PALI-X-55Bmode=per-task finetuning2023.07 | 126.3 | — | |
| InstructBLIPzero-shot=true, backbone=Vicuna-7B2023.05 | 123.1 | — | |
| InstructBLIPBackbone LLM=Vicuna-7B2024.06 | 123.1 | — | |
| InstructBLIPLLM=Vicuna-7B2023.12 | 123.1 | — | |
| InstructBLIPzero-shot=true, backbone=Vicuna-13B2023.05 | 121.9 | — | |
| InstructBLIPzero-shot=true, backbone=Vicuna-13B2023.08 | 121.9 | — | |
| InstructBLIPBackbone=Vicuna-13B, Evaluation protocol=0-shot2023.12 | 121.9 | — | |
| InstructBLIPBackbone LLM=Vicuna-13B2024.06 | 121.9 | — | |
| InstructBLIPLLM=Vicuna-13B2023.12 | 121.9 | — | |
| Qwen-VLzero-shot=true, backbone=Qwen-7B2023.08 | 121.4 | — | |
| Qwen-VL-Chatzero-shot=true2023.08 | 120.2 | — | |
| Qwen-VLBackbone=Qwen-7B, Evaluation protocol=0-shot2023.12 | 120.2 | — | |
| Qwen-VLBackbone LLM=Qwen-7B2024.06 | 120.2 | — | |
| Qwen-VLLLM=Qwen-7B2023.12 | 120.2 | — | |
| InstructBLIPzero-shot=true, backbone=FlanT5-XXL2023.05 | 120 | — | |
| InstructBLIPzero-shot=true, backbone=FlanT5-XL2023.05 | 119.9 | — | |
| MM1-30BShot=16, Model Scale=30B2024.03 | 116 | — | |
| MM1-30BShot=162024.03 | 116 | — | |
| OneLLM-7BBackbone LLM=LLaMA2-7B2024.06 | 115.9 | — | |
| OneLLM-7BLLM=LLAMA2-7B2023.12 | 115.9 | — | |
| ChatBridge-13BBackbone LLM=Vicuna-13B2024.06 | 115.7 | — | |
| ChatBridge-13BLLM=Vicuna-13B2023.12 | 115.7 | — | |
| MiCo-Chat-7BBackbone LLM=Vicuna-7B2024.06 | 111.8 | — | |
| MM1-30BShot=8, Model Scale=30B2024.03 | 111.6 | — | |
| MM1-30BShot=82024.03 | 111.6 | — | |
| MM1-7BShot=162024.03 | 111.1 | — | |
| SimVLMModel Size=Huge2021.08 | 110.3 | 14.5 | |
| Emu-Izero-shot=true, instruction_tuned=true2023.07 | 108.8 | — | |
| SimVLMModel Size=Large2021.08 | 108.5 | 14.2 | |
| MM1-3BShot=162024.03 | 107.6 | — | |
| BLIP-2zero-shot=true, backbone=Vicuna-7B2023.05 | 107.5 | — | |
| BLIP-2Backbone LLM=Vicuna-7B2024.06 | 107.5 | — | |
| BLIP-2LLM=Vicuna-7B2023.12 | 107.5 | — | |
| IDEFICS-80BShot=16, Model Scale=80B2024.03 | 107 | — | |
| IDEFICS-80BShot=162024.03 | 107 | — | |
| MM1-7BShot=8, Model Scale=7B2024.03 | 106.6 | — | |
| MM1-7BShot=82024.03 | 106.6 | — | |
| IDEFICS-80BShot=8, Model Scale=80B2024.03 | 105.7 | — | |
| IDEFICS-80BShot=82024.03 | 105.7 | — | |
| MM1-3BShot=8, Model Scale=3B2024.03 | 104.7 | — | |
| MM1-3BShot=82024.03 | 104.7 | — | |
| BLIP-2zero-shot=true, backbone=FlanT5-XL2023.05 | 104.5 | — | |
| BLIP-2zero-shot=true, backbone=Vicuna-13B2023.05 | 103.9 | — | |
| BLIP-2zero-shot=true, backbone=Vicuna-13B2023.08 | 103.9 | — | |
| BLIP-2Backbone=Vicuna-13B, Evaluation protocol=0-shot2023.12 | 103.9 | — | |
| BLIP-2Backbone LLM=Vicuna-13B2024.06 | 103.9 | — | |
| BLIP-2LLM=Vicuna-13B2023.12 | 103.9 | — | |
| MM1-30BShot=42024.03 | 103.8 | — | |
| AIMv2architecture=ViT-3B/14, #patches=576, resolution=336px2024.11 | 102.9 | — | |
| AIMv2architecture=ViT-H/14, #patches=576, resolution=336px2024.11 | 102.7 | — | |
| OriginalBit=162025.03 | 102.5 | — | |
| QuIP#Bit=22025.03 | 102.1 | — | |
| CASP_QuIP#Bit=22025.03 | 102.1 | — | |
| AIMv2architecture=ViT-1B/14, #patches=576, resolution=336px2024.11 | 101.5 | — | |
| SimVLMhugezero-shot=true2021.10 | 101.4 | — | |
| Unified-IO-XLzero-shot=true2023.08 | 100 | — | |
| AIMv2architecture=ViT-L/14, #patches=576, resolution=336px2024.11 | 99.9 | — | |
| MM1-3BShot=42024.03 | 99.7 | — | |
| IDEFICS-80BShot=42024.03 | 99.6 | — | |
| BLIP-2zero-shot=true, backbone=FlanT5-XXL2023.05 | 98.4 | — | |
| Emuzero-shot=true, prompt_strategy=Chain-of-Thought2023.07 | 96.5 | — | |
| MM1-7BShot=42024.03 | 96.2 | — | |
| SimVLMModel Size=Base2021.08 | 94.8 | 13.1 | |
| SigLIParchitecture=ViT-So400M/14, #patches=7522024.11 | 94.5 | — | |
| OpenAI CLIParchitecture=ViT-L/14, #patches=5762024.11 | 93.3 | — | |
| SigLIParchitecture=ViT-L/14, #patches=5762024.11 | 92.7 | — | |
| VinVLModel Size=Large2021.08 | 92.5 | 13.1 | |
| CASP_GPTQBit=2.22025.03 | 92.4 | — | |
| CASP_AQLMBit=22025.03 | 91.5 | — | |
| IDEFICS-9BShot=162024.03 | 89.4 | — | |
| DINOv2architecture=ViT-g/14, #patches=30342024.11 | 89.1 | — | |
| GenLIPArch=g/16, Data=8.0B2026.05 | 88.3 | — | |
| GenLIPArch=So/16, Data=8.0B2026.05 | 87.5 | — | |
| IDEFICS-9BShot=8, Model Scale=9B2024.03 | 86.8 | — | |
| IDEFICS-9BShot=82024.03 | 86.8 | — | |
| SigLIP2Arch=g/16, Data=40.0B2026.05 | 84.4 | — | |
| OVision2Arch=L/16, Data=12.8B2026.05 | 84.3 | — | |
| SigLIP2Arch=So/16, Data=40.0B2026.05 | 84.3 | — | |
| SigLIPArch=L/16, Data=40.0B2026.05 | 84 | — | |
| SigLIP2Arch=L/16, Data=40.0B2026.05 | 82.9 | — | |
| GenLIPArch=L/16, Data=8.0B2026.05 | 82.6 | — | |
| IDEFICS-9BShot=42024.03 | 81.3 | — | |
| OSCARLModel scale=Large2020.04 | 80.9 | 11.3 | |
| UNITERModel Size=Large2021.08 | 80.9 | 11.3 | |
| AIMv2Arch=L/14, Data=12.0B2026.05 | 80.1 | — | |
| OSCARBModel scale=Base2020.04 | 78.8 | 11.7 | |
| SoTABModel scale=Base2020.04 | 73.1 | 11.2 | |
| AQLMBit=22025.03 | 73.1 | — | |
| IDEFICS-80BBackbone=LLaMA-65B, Evaluation protocol=0-shot2023.12 | 65 | — | |
| IDEFICS-80BShot=0, Text-only demonstrations=True, Model Scale=80B2024.03 | 65 | — | |
| IDEFICS-80BShot=0, Includes text-only demonstrations=true2024.03 | 65 | — | |
| SoTASModel scale=Small2020.04 | 61.5 | 9.2 | |
| MM1-7BShot=0, Model Scale=7B2024.03 | 61 | — | |
| MM1-7BShot=02024.03 | 61 | — | |
| MetaLMzero-shot=true2023.07 | 58.7 | — | |
| MM1-3BShot=0, Model Scale=3B2024.03 | 55.6 | — | |
| MM1-3BShot=02024.03 | 55.6 | — | |
| CLIPArch=L/14, Data=12.8B2026.05 | 55.5 | — |