File size: 7,079 Bytes
cde8217 255a4d2 a7b9847 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 |
Model,Language Model,Model Size,Evaluation Method,Avg. Single,Avg. Multi,Avg. Video,Avg. P1,Avg. P2,Avg. P3,Scene Understanding,Instance Identity,Instance Attribute,Instance Location,Instance Counting,Spatial Relation,Instance Interaction,Visual Reasoning,Text Recognition,Celebrity Recognition,Landmark Recognition,Chart Understanding,Visual Referring Expression,Science Knowledge,Emotion Recognition,Visual Mathematics,Difference Spotting,Meme Comprehension,Global Video Understanding,Action Recognition,Action Predicion,Procedure Understanding,In-Context Captioning,Interleaved Image-Text Analysis,Text-to-Image Generation,Next Image Prediction,Text-Image Creation
[BLIP-2](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,44.1,28.2,34.8,41.0,35.3,0.0,58.5,48.6,49.0,39.1,43.4,36.2,48.5,52.9,60.7,51.8,51.4,19.2,43.2,52.4,29.3,22.0,17.8,38.6,42.5,37.7,36.2,22.9,40.0,30.6,0.0,0.0,0.0
[InstructBLIP](https://github.com/salesforce/LAVIS),Flan-T5-XL,3B,PPL,45.5,29.0,35.7,42.2,35.7,0.0,58.9,49.7,61.7,35.1,58.1,34.9,47.4,55.9,61.4,48.5,45.4,26.4,41.7,47.7,34.5,21.2,22.8,35.2,41.5,36.1,40.5,24.5,36.7,34.7,0.0,0.0,0.0
[InstructBLIP-Vicuna](https://github.com/salesforce/LAVIS),Vicuna-7B,7B,PPL,43.4,46.0,31.5,41.4,29.7,0.0,53.6,43.9,49.0,37.8,56.5,35.8,43.3,56.2,57.2,60.3,44.4,27.9,39.2,39.4,23.0,26.5,36.5,55.4,40.4,38.6,31.2,15.6,26.7,32.7,0.0,0.0,0.0
[LLaVA](https://github.com/haotian-liu/LLaVA),LLaMA-7B,7B,PPL,40.6,38.5,31.0,38.7,30.2,0.0,53.8,47.5,38.3,34.2,42.0,34.7,40.2,52.9,46.4,51.8,45.6,30.3,40.2,37.6,34.3,20.5,27.0,50.0,44.1,36.2,25.1,18.6,40.0,20.4,0.0,0.0,0.0
[MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4),Vicuna-7B,7B,PPL,41.7,32.9,33.4,39.4,34.1,0.0,56.3,49.2,45.8,37.9,45.3,32.6,47.4,57.1,41.8,55.2,45.2,20.2,41.2,43.3,24.2,25.0,19.0,46.7,39.0,38.7,27.4,28.6,45.8,22.5,0.0,0.0,0.0
[VPGTrans](https://github.com/VPGTrans/VPGTrans),LLaMA-7B,7B,PPL,37.4,34.3,32.6,36.2,23.9,0.0,46.9,38.6,33.6,35.6,27.5,34.4,33.0,50.8,47.6,52.4,38.2,30.1,34.7,36.1,31.5,27.3,24.6,44.0,37.8,38.2,20.9,33.5,19.2,28.6,0.0,0.0,0.0
[MultiModal-GPT](https://github.com/open-mmlab/Multimodal-GPT),LLaMA-7B,7B,PPL,37.5,48.3,31.5,37.4,34.9,0.0,46.9,42.5,32.0,32.3,27.7,29.7,29.9,48.3,35.2,60.9,50.4,24.2,42.2,37.6,32.1,27.3,40.1,56.5,37.6,38.7,25.3,24.4,39.2,30.6,0.0,0.0,0.0
[Otter](https://github.com/Luodian/Otter),LLaMA-7B,7B,PPL,37.6,37.1,31.3,36.4,36.6,0.0,45.9,39.7,31.9,31.6,26.4,32.0,33.0,49.2,39.3,59.7,53.0,23.6,41.2,36.1,37.3,22.0,27.4,46.7,36.6,37.9,26.0,24.8,42.5,30.6,0.0,0.0,0.0
[OpenFlamingo](https://github.com/mlfoundations/open_flamingo),LLaMA-7B,7B,PPL,37.5,47.4,31.3,37.3,35.5,0.0,46.7,42.3,31.7,33.4,27.4,29.8,29.9,47.7,35.6,60.3,49.8,24.2,42.2,39.0,32.1,27.3,39.9,54.9,37.6,38.4,25.2,24.1,38.3,32.7,0.0,0.0,0.0
[LLaMA-AdapterV2](https://github.com/OpenGVLab/LLaMA-Adapter),LLaMA-7B,7B,PPL,39.0,40.7,29.8,37.5,0.0,0.0,45.2,38.5,29.3,33.0,29.7,35.5,39.2,52.0,48.7,58.5,46.4,24.2,41.2,40.1,39.7,23.5,29.1,52.2,41.9,38.2,18.8,20.3,0.0,0.0,0.0,0.0,0.0
[GVT](https://github.com/TencentARC/GVT),Vicuna-7B,7B,PPL,33.5,50.4,30.1,34.4,38.6,0.0,41.7,35.5,31.8,29.5,36.2,32.0,32.0,51.1,35.2,39.4,36.4,25.0,36.2,31.1,20.6,22.7,41.5,59.2,40.4,29.7,26.3,24.1,42.5,34.7,0.0,0.0,0.0
[mPLUG-Owl](https://github.com/X-PLUG/mPLUG-Owl),LLaMA-7B,7B,PPL,41.3,44.2,29.4,39.4,28.9,0.0,49.7,45.3,32.5,36.7,27.3,32.7,44.3,54.7,49.2,70.9,49.6,23.2,44.2,44.0,32.5,23.5,33.5,54.9,42.0,37.8,18.3,19.3,29.2,28.6,0.0,0.0,0.0
[Kosmos-2](https://github.com/microsoft/unilm/tree/master/kosmos-2),Decoder only 1.3B,1.3B,PPL,49.5,34.0,39.7,46.3,23.3,0.0,63.4,57.1,58.5,44.0,41.4,37.9,55.7,60.7,68.1,82.1,51.4,21.2,48.2,43.7,30.7,28.0,25.2,42.8,48.5,40.8,39.5,30.0,24.2,22.5,0.0,0.0,0.0
[Qwen-VL-Chat](https://huggingface.co/Qwen/Qwen-VL-Chat),Qwen-7B,7B,PPL,46.0,40.8,32.8,43.1,35.5,0.0,56.5,47.6,54.8,46.9,54.2,40.3,55.7,55.0,47.4,62.4,55.6,25.2,43.7,41.2,20.6,28.8,34.3,47.2,39.7,42.8,29.6,19.1,42.5,28.6,0.0,0.0,0.0
[Qwen-VL-plus](https://github.com/QwenLM/Qwen-VL/tree/master?tab=readme-ov-file#qwen-vl-plus),Qwen-LM,-,PPL for A/B/C/D,69.0,74.3,45.9,65.3,30.5,0.0,76.6,77.7,76.3,65.1,65.8,55.9,73.2,77.9,61.8,97.0,97.2,39.5,73.4,75.8,51.7,38.6,66.7,81.8,51.8,54.5,29.3,48.0,28.3,32.7,0.0,0.0,0.0
[LLaVA-1.5](https://github.com/haotian-liu/LLaVA),vicuna-7B,7B,PPL,50.8,43.0,35.8,47.3,30.8,0.0,63.7,62.4,66.7,51.3,60.2,38.5,47.4,59.8,69.0,60.6,49.8,25.0,45.7,56.7,31.1,24.2,35.7,50.3,46.1,39.4,29.4,28.1,39.2,22.5,0.0,0.0,0.0
[IDEFICS-9b-instruct](https://huggingface.co/HuggingFaceM4/idefics-9b-instruct),LLaMA-7B,7B,PPL,37.9,52.5,31.4,38.0,40.3,0.0,48.2,38.2,37.8,32.9,29.0,32.4,37.1,54.1,45.5,52.4,52.8,22.6,42.7,33.2,26.6,21.2,56.5,48.4,42.7,38.6,23.6,20.5,45.8,34.7,0.0,0.0,0.0
[InternLM-XComposer-VL](https://github.com/InternLM/InternLM-XComposer),InternLM-7B,7B,PPL,64.2,52.2,42.8,59.2,32.1,0.0,74.8,70.5,67.6,60.5,55.3,53.4,76.3,76.1,61.4,86.1,78.0,27.2,60.3,84.8,68.9,25.8,47.7,56.6,58.6,49.9,37.6,24.9,27.5,36.7,0.0,0.0,0.0
[SPHINXv1-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,63.0,39.7,38.4,56.4,31.9,0.0,75.5,72.4,75.1,63.1,67.6,50.4,64.9,76.7,60.0,81.5,82.4,21.8,60.3,58.5,65.5,32.6,35.9,43.4,52.4,41.2,33.9,26.1,33.3,30.6,0.0,0.0,0.0
[SPHINXv2-1k](https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX),LLaMA-2-13B,13B,Generate,66.7,43.9,42.6,60.2,37.0,0.0,77.5,78.5,76.6,69.0,71.0,57.5,73.2,77.6,62.1,82.7,85.2,44.5,62.3,60.3,65.3,23.5,45.7,42.1,54.6,48.1,37.9,29.9,33.3,40.8,0.0,0.0,0.0
[Emu](https://github.com/baaivision/Emu),LLaMA-13B,13B,PPL,45.3,33.2,36.1,42.5,41.1,41.4,59.0,50.0,43.7,37.1,44.3,33.6,49.5,58.3,61.4,68.8,61.6,19.0,45.7,41.5,24.2,26.4,29.3,37.1,41.9,42.7,37.9,21.8,51.7,30.6,46.8,43.2,34.2
[Next-GPT](https://github.com/NExT-GPT/NExT-GPT),vicuna-7B,7B,PPL,30.8,31.6,29.9,30.7,35.6,33.9,36.4,35.1,25.6,29.9,36.1,30.9,39.2,41.7,31.0,30.9,27.4,21.2,34.2,31.8,24.4,17.4,24.2,39.0,35.5,33.8,25.6,24.5,46.7,24.5,45.1,19.8,36.7
[SEED-LLaMA](https://github.com/AILab-CVC/SEED),LLaMA2-Chat-13B,13B,PPL,46.5,35.5,37.6,43.9,43.4,52.3,64.0,55.0,51.3,45.4,43.3,37.9,56.7,59.2,57.0,55.5,52.8,18.8,49.3,44.8,28.8,24.4,29.5,41.5,46.7,39.4,43.9,20.3,54.2,32.7,50.2,40.7,65.8
[GPT-4V](https://openai.com/research/gpt-4v-system-card),\-,-,Generate,70.0,78.6,61.3,69.2,44.2,0.0,77.5,73.9,70.6,61.8,56.8,56.9,74.2,78.5,82.3,91.8,97.4,45.1,71.9,66.1,71.1,43.9,67.9,89.3,64.5,65.7,51.7,63.4,29.2,59.2,0.0,0.0,0.0
[VideoChat](https://github.com/OpenGVLab/Ask-Anything),Vicuna-7B,7B,PPL,37.4,41.0,33.4,37.0,35.3,0.0,44.3,40.7,32.2,36.9,32.9,32.6,42.3,51.1,45.7,35.2,46.8,20.6,43.2,39.4,34.3,19.7,30.3,51.6,41.5,34.0,30.6,27.4,40.0,30.6,0.0,0.0,0.0
[Video-ChatGPT](https://github.com/mbzuai-oryx/Video-ChatGPT),LLaMA-7B,7B,PPL,35.8,53.8,30.2,36.4,31.0,0.0,44.1,37.0,35.8,30.7,44.2,31.1,29.9,49.9,39.8,49.7,40.6,22.0,33.2,37.2,22.4,25.0,46.1,61.4,42.6,32.2,27.0,19.0,37.5,24.5,0.0,0.0,0.0
[Valley](https://github.com/RupertLuo/Valley),LLaMA-13B,13B,PPL,34.9,44.7,28.0,34.5,32.2,0.0,45.3,36.4,33.7,30.6,27.1,31.5,35.1,52.0,35.2,44.9,43.4,23.8,33.2,37.2,26.0,22.7,37.1,52.2,31.5,32.1,21.9,26.5,35.8,28.6,0.0,0.0,0.0 |