Composed Video Retrieval on WebVid-CoVR (test)
5,982R@1CoVR-BLIP-2 (CA)
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CoVR-BLIP-2 (CA)Input modalities=Visual + Text, Fusion=CA, Backbone=BLIP-2, Finetuning=Finetuned on WebVid-CoVR2023.08 | 5,982 | 8,384 | 9,128 | 98.24 | — | |
| BLIP (CA)Input modalities=Visual + Text, Fusion=CA, Backbone=BLIP, Finetuning=Finetuned on WebVid-CoVR2023.08 | 5,595 | 8,122 | 8,905 | 98.08 | — | |
| CoVR-BLIP-2 (MLP)Input modalities=Visual + Text, Fusion=MLP, Backbone=BLIP-2, Finetuning=Finetuned on WebVid-CoVR2023.08 | 5,188 | 7,938 | 8,646 | 97.42 | — | |
| CLIPInput modalities=Visual + Text, Fusion=MLP, Backbone=CLIP, Finetuning=Finetuned on WebVid-CoVR2023.08 | 5,086 | 7,746 | 8,532 | 96.75 | — | |
| BLIPInput modalities=Visual + Text, Fusion=MLP, Backbone=BLIP, Finetuning=Finetuned on WebVid-CoVR2023.08 | 5,059 | 7,465 | 8,357 | 95.46 | — | |
| BLIP-2Input modalities=Visual + Text, Fusion=Avg, Backbone=BLIP-2, Finetuning=Not finetuned2023.08 | 4,566 | 7,171 | 8,130 | 94.8 | — | |
| BLIPInput modalities=Visual + Text, Fusion=Avg, Backbone=BLIP, Finetuning=Not finetuned2023.08 | 4,546 | 7,046 | 7,954 | 93.27 | — | |
| CLIPInput modalities=Visual + Text, Fusion=Avg, Backbone=CLIP, Finetuning=Not finetuned2023.08 | 4,437 | 6,913 | 7,762 | 93 | — | |
| BLIP-2Input modalities=Visual, Fusion=none, Backbone=BLIP-2, Finetuning=Finetuned on WebVid-CoVR2023.08 | 3,603 | 6,424 | 7,477 | 92.64 | — | |
| BLIP-2Input modalities=Visual, Fusion=none, Backbone=BLIP-2, Finetuning=Not finetuned2023.08 | 3,310 | 5,955 | 6,980 | 88.85 | — | |
| BLIP-2Input modalities=Text, Fusion=none, Backbone=BLIP-2, Finetuning=Finetuned on WebVid-CoVR2023.08 | 2,332 | 4,621 | 5,622 | 78.36 | — | |
| BLIP-2Input modalities=Text, Fusion=none, Backbone=BLIP-2, Finetuning=Not finetuned2023.08 | 1,874 | 3,811 | 4,797 | 67.21 | — | |
| PREGEN2026.01 | 99.73 | 99.92 | 99.96 | 100 | — | |
| ReCoVRTurn=52026.05 | 90.22 | 94.13 | 96.24 | 99.57 | 95.04 | |
| ReCoVRTurn=42026.05 | 87.72 | 93.04 | 95.27 | 99.3 | 93.83 | |
| ReCoVRTurn=32026.05 | 84.39 | 91.28 | 94.56 | 99.18 | 92.35 | |
| ReCoVRTurn=22026.05 | 80.32 | 89.28 | 93.81 | 98.9 | 90.45 | |
| ReCoVRTurn=12026.05 | 74.3 | 87.28 | 93.08 | 98.75 | 88.35 | |
| UNITEinstructNumber of parameters=7B2026.01 | 72.5 | 90.8 | 95.3 | 99.5 | — | |
| UniCVR (Stage II)Zero-Shot=true2026.04 | 66.77 | 86.31 | 91.88 | 98.29 | 85.81 | |
| X-AlignerTraining=Finetuned, Backbone=BLIP-2, Fusion=X-Aligner2026.01 | 63.93 | 87.01 | 92.41 | — | — | |
| ReTrackModel Category=CVR Models2026.04 | 63.85 | 87.05 | 92.8 | 99.1 | 85.7 | |
| Dense-CoVRTraining=Finetuned, Backbone=BLIP-2, Fusion=CA2026.01 | 63.8 | 87.5 | 92.4 | — | — | |
| CADPMode=Single-Turn2026.05 | 63.62 | 87.36 | 93.19 | 99.3 | 85.87 | |
| IMAGINEModel Category=CVR Models2026.06 | 63.51 | 87.26 | 92.72 | 99.03 | 85.63 | |
| X-AlignerTraining=Finetuned, Backbone=BLIP, Fusion=X-Aligner2026.01 | 63.5 | 85.95 | 91.59 | — | — | |
| HUDZero-Shot=false2026.04 | 63.38 | 86.93 | 92.29 | 98.76 | 85.34 | |
| HUDMode=Single-Turn2026.05 | 63.38 | 86.93 | 92.29 | 98.76 | 85.34 | |
| UniCVR (Stage I)Zero-Shot=true2026.04 | 62.96 | 83.67 | 90.37 | 97.81 | 83.7 | |
| X-Aligner (Stage 1)Training=Finetuned, Backbone=BLIP-2, Fusion=X-Aligner2026.01 | 62.79 | 86.54 | 92.06 | — | — | |
| REFINEMode=Single-Turn2026.05 | 62.48 | 86.82 | 92.41 | 98.87 | 85.15 | |
| X-Aligner (Stage 1)Training=Finetuned, Backbone=BLIP, Fusion=X-Aligner2026.01 | 61.89 | 84.55 | 90.92 | — | — | |
| WebVid-CoVRTraining=Finetuned, Backbone=BLIP-2, Fusion=CA2026.01 | 60.4 | 84.5 | 91.4 | — | — | |
| ECDETraining=Finetuned, Backbone=BLIP, Fusion=CA2026.01 | 60.12 | 84.32 | 91.27 | — | — | |
| CoVR_EnrichModel Category=CVR Models2026.04 | 60.12 | 84.32 | 91.27 | 98.72 | 83.61 | |
| CoVR_EnrichZero-Shot=false2026.04 | 60.12 | 84.32 | 91.27 | 98.72 | 83.61 | |
| ECDEMode=Single-Turn2026.05 | 60.12 | 84.32 | 91.27 | 98.72 | 83.61 | |
| CoVR_EnrichModel Category=CVR Models2026.06 | 60.12 | 84.32 | 91.27 | 98.72 | 83.61 | |
| ECDE2026.01 | 60.1 | 84.3 | 91.3 | 98.7 | — | |
| CoVR-2Training=Finetuned, Backbone=BLIP-2, Fusion=CA2026.01 | 59.82 | 83.84 | 91.28 | — | — | |
| CoVR-2Model Category=CVR Models2026.04 | 59.82 | 83.84 | 91.28 | 98.24 | 83.3 | |
| CoVR-2Zero-Shot=false2026.04 | 59.82 | 83.84 | 91.28 | 98.24 | 83.3 | |
| CoVR-2Mode=Single-Turn2026.05 | 59.82 | 83.84 | 91.28 | 98.24 | 83.3 | |
| CoVR-2Model Category=CVR Models2026.06 | 59.82 | 83.84 | 91.28 | 98.24 | 83.3 | |
| CoVR-22026.01 | 59.8 | 83.8 | 91.3 | 98.2 | — | |
| Ventura et al.Modalities=V+T, Backbone=BLIP2, F=15, Learning Mode=Fine-tuned2025.12 | 59.8 | 83.8 | 91.3 | 98.2 | — | |
| CoVRTraining=Finetuned, Backbone=BLIP, Fusion=CA2026.01 | 55.95 | 81.22 | 89.05 | — | — | |
| FDCAModel Category=CVR Models2026.04 | 54.8 | 82.27 | 89.84 | 97.7 | 81.15 | |
| FDCAZero-Shot=false2026.04 | 54.8 | 82.27 | 89.84 | 97.7 | 81.15 | |
| FDCAMode=Single-Turn2026.05 | 54.8 | 82.27 | 89.84 | 97.7 | 81.15 | |
| FDCAModel Category=CVR Models2026.06 | 54.8 | 82.27 | 89.84 | 97.7 | 81.15 | |
| ReCoVRTurn=02026.05 | 54.3 | 77 | 84.35 | 95.77 | 77.86 | |
| CoVRModel Category=CVR Models2026.04 | 53.13 | 79.93 | 86.85 | 97.69 | 79.4 | |
| CoVRZero-Shot=false2026.04 | 53.13 | 79.93 | 86.85 | 97.69 | 79.4 | |
| CoVRMode=Single-Turn2026.05 | 53.13 | 79.93 | 86.85 | 97.69 | 79.4 | |
| CoVRModel Category=CVR Models2026.06 | 53.13 | 79.93 | 86.85 | 97.69 | 79.4 | |
| CoVR2026.01 | 53.1 | 79.9 | 86.9 | 97.7 | — | |
| TARA w/ caps.Modalities=V+T, Backbone=Tarsier, F=15, Learning Mode=Zero-shot, use original image caption as prompt=true2025.12 | 53.1 | 78.6 | 86.4 | 97.3 | — | |
| Ventura et al.Modalities=V+T, Backbone=BLIP, F=15, Learning Mode=Fine-tuned2025.12 | 53.1 | 79.9 | 86.9 | 97.7 | — | |
| BLIPModalities=V+T, Backbone=BLIP, F=1, Learning Mode=Fine-tuned, Version=Secondary Entry2025.12 | 51.8 | 78.3 | 85.8 | 97.1 | — | |
| TFR-CVRMode=Single-Turn2026.05 | 51.7 | 75.3 | 80.7 | — | — | |
| TARAModalities=V+T, Backbone=Tarsier, F=8, Learning Mode=Zero-shot2025.12 | 50.8 | 75.6 | 83 | 96 | — | |
| CLIPModalities=V+T, Backbone=CLIP, F=1, Learning Mode=Fine-tuned2025.12 | 50.6 | 77.1 | 85.1 | 96.6 | — | |
| BLIPModalities=V+T, Backbone=BLIP, F=1, Learning Mode=Fine-tuned2025.12 | 50.6 | 74.8 | 83.4 | 95.5 | — | |
| Reason-then-Retrieve2026.03 | 49.15 | 70.72 | 79.25 | 0.9342 | — | |
| BSE-CoVR2026.03 | 48.08 | 73.36 | 81.06 | 0.9378 | — | |
| BLIP-2Training=Not finetuned, Backbone=BLIP-2, Fusion=Avg2026.01 | 45.66 | 71.71 | 81.3 | — | — | |
| CoVR-BLIP-22026.03 | 45.66 | 71.71 | 81.3 | 0.948 | — | |
| CoVR-2Zero-Shot=true2026.04 | 45.66 | 71.71 | 81.3 | 94.8 | 73.37 | |
| BLIPModalities=V+T, Backbone=BLIP, F=15, Learning Mode=Zero-shot2025.12 | 45.5 | 70.5 | 79.5 | 93.3 | — | |
| BLIPTraining=Not finetuned, Backbone=BLIP, Fusion=Avg2026.01 | 45.46 | 70.46 | 79.54 | — | — | |
| CoVR-BLIP2026.03 | 45.46 | 70.46 | 79.54 | 0.9327 | — | |
| BLIPModel Category=Pre-trained Models2026.04 | 45.46 | 70.46 | 79.54 | 93.27 | 72.18 | |
| CoVRZero-Shot=true2026.04 | 45.46 | 70.46 | 79.54 | 93.27 | 72.18 | |
| BLIPModel Category=Pre-trianed Models2026.06 | 45.46 | 70.46 | 79.54 | 93.27 | 72.18 | |
| CLIPModalities=V+T, Backbone=CLIP, F=15, Learning Mode=Zero-shot2025.12 | 44.4 | 69.1 | 77.6 | 93 | — | |
| CLIPTraining=Not finetuned, Backbone=CLIP, Fusion=Avg2026.01 | 44.37 | 69.13 | 77.62 | — | — | |
| CLIPModel Category=Pre-trained Models2026.04 | 44.37 | 69.13 | 77.62 | 93 | 71.03 | |
| CLIPModel Category=Pre-trianed Models2026.06 | 44.37 | 69.13 | 77.62 | 93 | 71.03 | |
| BLIPModalities=V, Backbone=BLIP, F=15, Learning Mode=Fine-tuned2025.12 | 38.9 | 65 | 74 | 92.1 | — | |
| BLIPModalities=V, Backbone=BLIP, F=15, Learning Mode=Zero-shot2025.12 | 34.9 | 59.2 | 68 | 86 | — | |
| BLIPModalities=T, Backbone=BLIP, Learning Mode=Fine-tuned2025.12 | 23.7 | 45.9 | 55.1 | 77 | — | |
| BLIPModalities=T, Backbone=BLIP, Learning Mode=Zero-shot2025.12 | 19.7 | 37.1 | 45.9 | 65.1 | — | |
| PREGENNumber of layers=12026.01 | 14.2 | 30.71 | 40.18 | 67.45 | — | |
| RandomInput modalities=N/A, Fusion=N/A, Backbone=N/A, Finetuning=none2023.08 | 8 | 23 | 35 | 1.76 | — | |
| Chance2025.12 | 0.1 | 0.2 | 0.4 | 1.8 | — |