Composed Image Retrieval on CIRR (test)
56.58Recall@1WRF4CIR
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| WRF4CIRBackbone=BLIP-2 ViT-G, Additional Data=false2026.04 | 56.58 | 85.45 | 92.24 | 98.68 | 80.48 | 92.6 | 96.84 | 82.96 | — | — | — | — | |
| ReCALL (R_refine)2026.02 | 55.52 | 84.07 | 91.83 | 98.55 | 81.49 | 93.35 | 97.64 | 82.81 | — | — | — | — | |
| MCoT-MVSType=VLP Model-Based2026.03 | 55.33 | 84.75 | 91.45 | 98.55 | — | — | — | 82.52 | — | — | — | — | |
| SPRC + SPNBackbone=BLIP-22024.04 | 55.06 | 83.83 | 90.87 | 98.29 | 81.54 | 92.65 | 97.04 | 82.69 | — | — | — | — | |
| SPRC+SPNBackbone=BLIP-2 ViT-G, Additional Data=true2026.04 | 55.06 | 83.83 | 90.87 | 98.29 | 81.54 | 92.65 | 97.04 | 82.69 | — | — | — | — | |
| FAFA2023.11 | 54.48 | 84.07 | 91.48 | — | 81.05 | — | 97.11 | 82.56 | — | — | — | — | |
| SPRCRe-ranking=true2023.10 | 54.15 | 83.01 | 90.39 | 98.17 | 82.31 | 92.68 | 96.87 | 82.66 | — | — | — | — | |
| TMCIR2025.04 | 54.12 | 84.27 | 91.06 | 98.43 | 82.64 | 92.45 | 96.77 | 82.66 | — | — | — | — | |
| DQE-CIR2026.03 | 54.05 | 84.17 | 91.58 | 98.68 | 80.14 | 92.39 | 97.32 | — | — | — | 82.16 | — | |
| RUNCBackbone=BLIP-2 ViT-G, Additional Data=false2026.04 | 53.81 | 83.47 | 91.11 | 98.22 | 80.87 | 92.36 | 96.94 | 82.17 | — | — | — | — | |
| CSMCIR2026.01 | 53.76 | 84.15 | 91.25 | 98.19 | 80.82 | 92.72 | 97.11 | — | — | — | 82.49 | — | |
| CIR-LVLM2025.04 | 53.64 | 83.76 | 90.6 | 97.93 | 79.12 | 92.33 | 96.67 | 81.44 | — | — | — | — | |
| CIR-LVLMVenue=AAAI’252026.02 | 53.64 | 83.76 | 90.6 | 97.93 | 79.12 | 92.33 | 96.67 | 81.44 | — | — | — | — | |
| CIRLVLMType=VLP Model-Based2026.03 | 53.64 | 83.76 | 90.6 | 97.93 | — | — | — | 81.48 | — | — | — | — | |
| CIR-LVLMBackbone=Qwen-7B, Additional Data=false2026.04 | 53.64 | 83.76 | 90.6 | 97.93 | 79.12 | 92.33 | 96.67 | 81.44 | — | — | — | — | |
| TMEVenue=CVPR’252026.02 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | 82.01 | — | — | — | — | |
| TME2026.01 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | — | — | — | 82.01 | — | |
| TMEBackbone=BLIP-2 ViT-G, Additional Data=false2026.04 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | 82.01 | — | — | — | — | |
| TMENoise=0%2026.04 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | — | — | — | 82.01 | — | |
| TMENoise=0%2026.04 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | — | — | — | 82.01 | — | |
| TMENoise=0%2026.04 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | — | — | — | 82.01 | — | |
| TMENoise level=0%2026.04 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | — | — | — | 82.01 | — | |
| TMENoise=0%2026.06 | 53.42 | 82.99 | 90.24 | 98.15 | 81.04 | 92.58 | 96.94 | 82.02 | — | — | — | — | |
| CCINVenue=CVPR’252026.02 | 53.41 | 84.05 | 91.17 | 98 | — | — | — | — | — | — | — | — | |
| CCIN2026.01 | 53.41 | 84.05 | 91.17 | 98 | — | — | — | — | — | — | — | — | |
| CCINType=VLP Model-Based2026.03 | 53.41 | 84.05 | 91.17 | 98 | — | — | — | 81.66 | — | — | — | — | |
| CCINBackbone=BLIP-2 ViT-G, Additional Data=false2026.04 | 53.41 | 84.05 | 91.17 | 98 | — | — | — | — | — | — | — | — | |
| INTENTNoise=0%2026.04 | 53.37 | 83.16 | 90.73 | 98.22 | 80.24 | 92.37 | 96.89 | — | — | — | 81.7 | — | |
| INTENTNoise=0%2026.04 | 53.37 | 83.16 | 90.73 | 98.22 | 80.24 | 92.37 | 96.89 | — | — | — | 81.7 | — | |
| INTENTNoise level=0%2026.04 | 53.37 | 83.16 | 90.73 | 98.22 | 80.24 | 92.37 | 96.89 | — | — | — | 81.7 | — | |
| RCLNoise=0%2026.04 | 53.16 | 82.41 | 90.12 | 98.34 | 79.57 | 92.02 | 96.87 | — | — | — | 80.99 | — | |
| RCLNoise=0%2026.04 | 53.16 | 82.41 | 90.12 | 98.34 | 79.57 | 92.02 | 96.87 | — | — | — | 80.99 | — | |
| RCLNoise=0%2026.06 | 53.16 | 82.41 | 90.12 | 98.34 | 79.57 | 92.02 | 96.87 | 80.99 | — | — | — | — | |
| ConeSepNoise level=0%2026.04 | 53.06 | 83.45 | 90.6 | 98.34 | 81.23 | 92.62 | 97.24 | — | — | — | 82.34 | — | |
| MELT2026.03 | 52.76 | 83.04 | 90.33 | 97.94 | 80.63 | 92.68 | 96.77 | 81.84 | — | — | — | — | |
| HABITNoise=0%2026.04 | 52.71 | 82.64 | 90.63 | 98.19 | 80.99 | 92.77 | 97 | — | — | — | 81.82 | — | |
| HABITNoise=0%2026.04 | 52.71 | 82.64 | 90.63 | 98.19 | 80.99 | 92.77 | 97 | — | — | — | 81.82 | — | |
| HABITNoise level=0%2026.04 | 52.71 | 82.64 | 90.63 | 98.19 | 80.99 | 92.77 | 97 | — | — | — | 81.82 | — | |
| COMBINERModel Category=VLP Model-Based Methods2026.06 | 52.6 | 82.51 | 90.12 | 98.17 | 81.33 | 93.06 | 97.23 | — | — | — | 81.92 | — | |
| COMBINERNotes=Dynamic prototypes (no explicit labels required)2026.06 | 52.6 | — | — | — | 81.33 | — | — | — | — | — | — | — | |
| ConeSepNoise level=20%2026.04 | 52.29 | 82.19 | 89.98 | 98.19 | 78.66 | 91.76 | 96.75 | — | — | — | 80.43 | — | |
| QuReVenue=ICML’252026.02 | 52.22 | 82.53 | 90.31 | 98.17 | 78.51 | 91.28 | 96.48 | 80.52 | — | — | — | — | |
| QuRe2026.01 | 52.22 | 82.53 | 90.31 | 98.17 | 78.51 | 91.28 | 96.48 | — | — | — | 80.52 | — | |
| QuRe2026.03 | 52.22 | 82.53 | 90.31 | 98.17 | 78.51 | 91.28 | 96.48 | — | — | — | 80.52 | — | |
| QuRe2026.03 | 52.22 | 82.53 | 90.31 | 98.17 | 78.51 | 91.28 | 96.48 | 80.52 | — | — | — | — | |
| QuReNoise=0%2026.04 | 52.22 | 82.53 | 90.31 | 98.17 | 78.51 | 91.28 | 96.48 | — | — | — | 80.52 | — | |
| OFFSETCategory=VLP Model-Based Methods2025.07 | 52.19 | 82.6 | 90.07 | 98.07 | 81.37 | 93.08 | 97.54 | — | — | — | 81.99 | — | |
| Air-KnowNoise=0%2026.04 | 51.97 | 82.25 | 90.42 | 98.25 | 79.21 | 91.94 | 96.87 | — | — | — | 80.73 | — | |
| SPRC2023.10 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | 81.39 | — | — | — | — | |
| SPRCBackbone=BLIP-22024.04 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | 81.39 | — | — | — | — | |
| SPRC2025.04 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | 81.39 | — | — | — | — | |
| SPRCRef.=ICLR242023.11 | 51.96 | 82.12 | 89.74 | — | 80.65 | — | 96.6 | 81.39 | — | — | — | — | |
| SPRCVenue=ICLR’242026.02 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | 81.39 | — | — | — | — | |
| SPRC2026.01 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | — | — | — | 81.39 | — | |
| SPRCType=VLP Model-Based2026.03 | 51.96 | 82.12 | 89.74 | 97.18 | — | — | — | 80.25 | — | — | — | — | |
| SPRCCategory=VLP Model-Based Methods2025.07 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | — | — | — | 81.39 | — | |
| SPRC2026.03 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | 81.39 | — | — | — | — | |
| SPRCBackbone=BLIP-2 ViT-G, Additional Data=false2026.04 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | 81.39 | — | — | — | — | |
| SPRCNoise=0%2026.04 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | — | — | — | 81.39 | — | |
| SPRCNoise=0%2026.04 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | — | — | — | 81.39 | — | |
| SPRCNoise=0%2026.04 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | — | — | — | 81.39 | — | |
| SPRCNoise level=0%2026.04 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | — | — | — | 81.39 | — | |
| SPRCModel Category=VLP Model-Based Methods2026.06 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | — | — | — | 81.39 | — | |
| SPRCNoise=0%2026.06 | 51.96 | 82.12 | 89.74 | 97.69 | 80.65 | 92.31 | 96.6 | 81.39 | — | — | — | — | |
| RankVRNoise=0%2026.06 | 51.94 | 81.89 | 90.41 | 97.41 | 80.12 | 92.4 | 95.65 | 81.01 | — | — | — | — | |
| RDENoise=0%2026.04 | 51.81 | 82.02 | 90.6 | 97.93 | 78.17 | 91.9 | 96.7 | — | — | — | 80.1 | — | |
| RDENoise=0%2026.04 | 51.81 | 82.02 | 90.6 | 97.93 | 78.17 | 91.9 | 96.7 | — | — | — | 80.1 | — | |
| RDENoise=0%2026.06 | 51.81 | 82.02 | 90.6 | 97.93 | 78.17 | 91.9 | 96.7 | 80.1 | — | — | — | — | |
| RankVRNoise=20%2026.06 | 51.72 | 81.72 | 89.05 | 98.02 | 78.77 | 91.89 | 96.48 | 80.25 | — | — | — | — | |
| HABITNoise=20%2026.04 | 51.68 | 81.02 | 89.24 | 97.81 | 78.2 | 91.66 | 96.75 | — | — | — | 79.61 | — | |
| HABITNoise=20%2026.04 | 51.68 | 81.02 | 89.24 | 97.81 | 78.2 | 91.66 | 96.75 | — | — | — | 79.61 | — | |
| HABITNoise level=20%2026.04 | 51.68 | 81.02 | 89.24 | 97.81 | 78.2 | 91.66 | 96.75 | — | — | — | 79.61 | — | |
| Air-KnowNoise=20%2026.04 | 51.37 | 81.73 | 89.78 | 98.07 | 79.06 | 91.66 | 96.54 | — | — | — | 80.4 | — | |
| TMENoise=20%2026.04 | 51.35 | 81.01 | 88.53 | 97.81 | 78.46 | 91.25 | 96.39 | — | — | — | 79.74 | — | |
| TMENoise=20%2026.04 | 51.35 | 81.01 | 88.53 | 97.81 | 78.46 | 91.25 | 96.39 | — | — | — | 79.74 | — | |
| TMENoise=20%2026.04 | 51.35 | 81.01 | 88.53 | 97.81 | 78.46 | 91.25 | 96.39 | — | — | — | 79.74 | — | |
| TMENoise level=20%2026.04 | 51.35 | 81.01 | 88.53 | 97.81 | 78.46 | 91.25 | 96.39 | — | — | — | 79.74 | — | |
| TMENoise=20%2026.06 | 51.35 | 81.01 | 88.53 | 97.81 | 78.46 | 91.25 | 96.39 | 79.74 | — | — | — | — | |
| ConeSepNoise level=50%2026.04 | 51.28 | 80.46 | 88.46 | 97.49 | 77.04 | 91.04 | 96.6 | — | — | — | 78.75 | — | |
| INTENTNoise=20%2026.04 | 51.25 | 81.36 | 90.02 | 98.05 | 77.95 | 91.4 | 96.46 | — | — | — | 79.66 | — | |
| INTENTNoise=20%2026.04 | 51.25 | 81.36 | 90.02 | 98.05 | 77.95 | 91.4 | 96.46 | — | — | — | 79.66 | — | |
| INTENTNoise level=20%2026.04 | 51.25 | 81.36 | 90.02 | 98.05 | 77.95 | 91.4 | 96.46 | — | — | — | 79.66 | — | |
| Baseline(R_base)2026.02 | 51.23 | 82.15 | 90.2 | 98.2 | 77.57 | 91.83 | 96.34 | 79.86 | — | — | — | — | |
| HUG2026.01 | 51.09 | 83.2 | 92.03 | 97.89 | 80.65 | 91.8 | 95.93 | — | — | — | 81.93 | — | |
| OursEvaluation Protocol=Transfer Learning, Pretrain Data=WebVid-CoVR, Variant=Optimal2024.03 | 51.03 | — | 88.93 | 97.53 | 76.51 | — | 95.76 | — | — | — | — | — | |
| VDGExtra data usage=true, LLM-based=true2026.01 | 50.96 | 80.15 | 86.86 | 94.46 | 77.45 | 90.65 | 96.1 | — | — | — | 78.8 | — | |
| Air-KnowNoise=50%2026.04 | 50.82 | 80.42 | 88.37 | 97.83 | 77.43 | 90.68 | 96.24 | — | — | — | 78.93 | — | |
| SPRC2026.03 | 50.75 | 80.58 | 88.72 | 97.59 | 79.57 | 91.76 | 96.7 | — | — | — | 80.07 | — | |
| Re-ranking2023.10 | 50.55 | 81.75 | 89.78 | 97.18 | 80.04 | 91.9 | 96.58 | 80.9 | — | — | — | — | |
| Ours R50Candidate re-ranking model=RK, top-K=502023.05 | 50.55 | 81.75 | 89.78 | 97.18 | 80.04 | 91.9 | 96.58 | 80.9 | — | — | — | — | |
| Re-ranking2025.04 | 50.55 | 81.75 | 89.78 | 97.18 | 80.04 | 94.29 | 96.8 | 80.9 | — | — | — | — | |
| Re-ranking2026.01 | 50.55 | 81.75 | 89.78 | 97.18 | 80.04 | 91.9 | 96.58 | — | — | — | 80.9 | — | |
| CandidateCategory=VLP Model-Based Methods2025.07 | 50.55 | 81.75 | 89.78 | 97.18 | 80.04 | 91.9 | 96.58 | — | — | — | 80.9 | — | |
| Re-rankingBackbone=BLIP ViT-B, Additional Data=false2026.04 | 50.55 | 81.75 | 89.78 | 97.18 | 80.04 | 91.9 | 96.58 | 80.9 | — | — | — | — | |
| CandidateModel Category=VLP Model-Based Methods2026.06 | 50.55 | 81.75 | 89.78 | 97.18 | 80.04 | 91.9 | 96.58 | — | — | — | 80.9 | — | |
| CoVR-2Venue=TPAMI’242026.02 | 50.43 | 81.08 | 88.89 | 98.05 | 76.75 | 90.34 | 95.78 | 79.28 | — | — | — | — | |
| CoVR-2Category=VLP Model-Based Methods2025.07 | 50.43 | 81.08 | 88.89 | 98.05 | 76.75 | 90.34 | 95.78 | — | — | — | 79.28 | — | |
| RCLNoise=20%2026.04 | 50.43 | 81.11 | 88.82 | 96.68 | 77.52 | 90.8 | 95.71 | — | — | — | 79.31 | — | |
| RCLNoise=20%2026.04 | 50.43 | 81.11 | 88.82 | 96.68 | 77.52 | 90.8 | 95.71 | — | — | — | 79.31 | — | |
| CoVR-2Model Category=VLP Model-Based Methods2026.06 | 50.43 | 81.08 | 88.89 | 98.05 | 76.75 | 90.34 | 95.78 | — | — | — | 79.28 | — |