Image Captioning on COCO full (val)
135.66CIDErHyper-ICL
Evaluation Results
| Method | Links | |
|---|---|---|
| Hyper-ICLModel=Idefics2-8b, Method Type=Learnable, # Params (M)=1.18 (×1)2026.06 | 135.66 | |
| MimICModel=Idefics2-8b, Method Type=Learnable, # Params (M)=0.26 (×0.22)2026.06 | 132.87 | |
| LIVEModel=Idefics2-8b, Method Type=Learnable, # Params (M)=0.13 (×0.11)2026.06 | 126.04 | |
| 8-shot ICLModel=Idefics2-8b, Method Type=Direct ICDs2026.06 | 122.51 | |
| Hyper-ICLModel=Idefics-9b, Method Type=Learnable, # Params (M)=1.18 (×1)2026.06 | 117.44 | |
| LoRAModel=Idefics2-8b, Method Type=Learnable, # Params (M)=17.6 (×14.9)2026.06 | 116.69 | |
| MimICModel=Idefics-9b, Method Type=Learnable, # Params (M)=0.26 (×0.22)2026.06 | 114.89 | |
| LIVEModel=Idefics-9b, Method Type=Learnable, # Params (M)=0.13 (×0.11)2026.06 | 112.76 | |
| RICESModel=Idefics2-8b, Method Type=Direct ICDs2026.06 | 111.44 | |
| RICESModel=Idefics-9b, Method Type=Direct ICDs2026.06 | 110.64 | |
| 32-shot ICLModel=Idefics-9b, Method Type=Direct ICDs2026.06 | 105.89 | |
| LoRAModel=Idefics-9b, Method Type=Learnable, # Params (M)=25.0 (×21.2)2026.06 | 97.75 | |
| TVModel=Idefics2-8b, Method Type=Non-Learnable2026.06 | 87.61 | |
| TVModel=Idefics-9b, Method Type=Non-Learnable2026.06 | 84.72 | |
| FVModel=Idefics2-8b, Method Type=Non-Learnable2026.06 | 75.24 | |
| FVModel=Idefics-9b, Method Type=Non-Learnable2026.06 | 74.01 | |
| Zero-shotModel=Idefics-9b, Method Type=Direct ICDs2026.06 | 63.06 | |
| Zero-shotModel=Idefics2-8b, Method Type=Direct ICDs2026.06 | 40 |