| CoLLM: A Large Language Model for Composed Image Retrieval | ✓ Link | 45.8 | | 84.7 | 95.8 | | CoLLM (finetuned - BLIP-L/16) | 2025-03-25 |
| CoVR-2: Automatic Data Construction for Composed Video Retrieval | ✓ Link | 43.74 | 73.61 | 83.95 | 96.1 | | CoVR-BLIP-2 | 2023-08-28 |
| ImageScope: Unifying Language-Guided Image Retrieval via Large Multimodal Model Collective Reasoning | ✓ Link | 39.37 | 67.54 | 78.05 | 92.94 | | ImageScope (CLIP-ViT-L/14) | 2025-03-13 |
| Imagine and Seek: Improving Composed Image Retrieval with an Imagined Proxy | | 39.25 | 70.07 | 80 | 94.89 | | IP-CIR + LDRE (CLIP G/14) | 2024-11-24 |
| Semantic Editing Increment Benefits Zero-Shot Composed Image Retrieval | ✓ Link | 38.87 | 69.42 | 79.42 | | | SEIZE (CLIP G/14) | 2024-10-28 |
| Zero-shot Composed Text-Image Retrieval | ✓ Link | 37.87 | 68.88 | | 93.86 | | TransAgg (Laion-CIR-Combined) | 2023-06-12 |
| Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval | ✓ Link | 37.26 | 67.25 | 77.33 | | | OSrCIR (CLIP G/14) | 2024-12-15 |
| SCOT: Self-Supervised Contrastive Pretraining For Zero-Shot Compositional Retrieval | | 36.82 | 64.34 | 74.48 | 93.42 | | SCOT (WACV 2025) | 2025-01-12 |
| CoLLM: A Large Language Model for Composed Image Retrieval | ✓ Link | 35.00 | | 78.6 | 94.2 | | CoLLM (Pretrained - BLIP-L/16) | 2025-03-25 |
| Vision-by-Language for Training-Free Compositional Image Retrieval | ✓ Link | 34.65 | 64.29 | | | | CIReVL (CLIP G/14) | 2023-10-13 |
| MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions | ✓ Link | 33.3 | 67.0 | 77.9 | 94.4 | | MagicLens (CoCa L) | 2024-03-28 |
| Training-free Zero-shot Composed Image Retrieval via Weighted Modality Fusion and Similarity | ✓ Link | 31.04 | 60.41 | | | | WeiMoCIR (CLIP G/14) | 2024-09-07 |
| Training-free Zero-shot Composed Image Retrieval via Weighted Modality Fusion and Similarity | ✓ Link | 30.94 | 60.87 | | | | WeiMoCIR (CLIP L/14) | 2024-09-07 |
| MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions | ✓ Link | 30.1 | 61.7 | 74.4 | 92.6 | | MagicLens (CLIP L) | 2024-03-28 |
| Imagine and Seek: Improving Composed Image Retrieval with an Imagined Proxy | | 29.76 | 58.82 | 71.21 | 90.41 | | IP-CIR + LDRE (CLIP L/14) | 2024-11-24 |
| CoLLM: A Large Language Model for Composed Image Retrieval | ✓ Link | 29.7 | | 72.8 | 91.5 | | CoLLM (Pretrained - CLIP-L/14) | 2025-03-25 |
| Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval | ✓ Link | 29.45 | 57.68 | 69.86 | | | OSrCIR (CLIP L/14) | 2024-12-15 |
| Training-free Zero-shot Composed Image Retrieval via Weighted Modality Fusion and Similarity | ✓ Link | 29.11 | 59.76 | | | | WeiMoCIR (CLIP H/14) | 2024-09-07 |
| Training-free Zero-shot Composed Image Retrieval via Weighted Modality Fusion and Similarity | ✓ Link | 26.31 | 57.69 | | | | WeiMoCIR (CLIP B/32) | 2024-09-07 |
| Reason-before-Retrieve: One-Stage Reflective Chain-of-Thoughts for Training-Free Zero-Shot Composed Image Retrieval | ✓ Link | 25.42 | 54.54 | 68.19 | | | OSrCIR (CLIP B/32) | 2024-12-15 |
| Vision-by-Language for Training-Free Compositional Image Retrieval | ✓ Link | 24.55 | 52.31 | | | | CIReVL (CLIP L/14) | 2023-10-13 |
| Vision-by-Language for Training-Free Compositional Image Retrieval | ✓ Link | 23.94 | 52.51 | | | | CIReVL (CLIP B/32) | 2023-10-13 |
| An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval | ✓ Link | | 67.47 | | | | RTD + LinCIR (CLIP G/14) | 2024-06-13 |
| LDRE: LLM-based Divergent Reasoning and Ensemble for Zero-Shot Composed Image Retrieval | ✓ Link | | 66.39 | | | | LDRE (CLIP G/14) | 2024-07-11 |
| Improving Composed Image Retrieval via Contrastive Learning with Scaling Positives and Negatives | ✓ Link | | 65.42 | | | 64.87 | SPN4CIR (SPN-CC) | 2024-04-17 |
| Language-only Efficient Training of Zero-shot Composed Image Retrieval | ✓ Link | | 64.72 | | | | LinCIR (CLIP G/14) | 2023-12-04 |
| MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions | ✓ Link | | 64.0 | | | | MagicLens (CoCa B) | 2024-03-28 |
| Pretrain like Your Inference: Masked Tuning Improves Zero-Shot Composed Image Retrieval | ✓ Link | | 58.87 | | | | MTCIR (BLIP B/16) | 2023-11-13 |
| MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions | ✓ Link | | 58.0 | | | | MagicLens (CLIP B) | 2024-03-28 |
| CompoDiff: Versatile Composed Image Retrieval With Latent Diffusion | ✓ Link | | 57.61 | | | | CompoDiff (CLIP G/14) | 2023-03-21 |
| Semantic Editing Increment Benefits Zero-Shot Composed Image Retrieval | ✓ Link | | 57.42 | | | | SEIZE (CLIP B/32) | 2024-10-28 |
| Semantic Editing Increment Benefits Zero-Shot Composed Image Retrieval | ✓ Link | | 57.16 | | | | SEIZE (CLIP L/14) | 2024-10-28 |
| An Efficient Post-hoc Framework for Reducing Task Discrepancy of Text Encoders for Composed Image Retrieval | ✓ Link | | 56.17 | | | | RTD + LinCIR (CLIP L/14) | 2024-06-13 |
| iSEARLE: Improving Textual Inversion for Zero-Shot Composed Image Retrieval | ✓ Link | | 55.69 | | | | iSEARLE (CLIP B/32) | 2024-05-05 |
| LDRE: LLM-based Divergent Reasoning and Ensemble for Zero-Shot Composed Image Retrieval | ✓ Link | | 55.57 | | | | LDRE (CLIP L/14) | 2024-07-11 |
| iSEARLE: Improving Textual Inversion for Zero-Shot Composed Image Retrieval | ✓ Link | | 55.18 | | | | iSEARLE-OTI (CLIP B/32) | 2024-05-05 |
| LDRE: LLM-based Divergent Reasoning and Ensemble for Zero-Shot Composed Image Retrieval | ✓ Link | | 55.13 | | | | LDRE (CLIP B/32) | 2024-07-11 |
| Context-I2W: Mapping Images to Context-dependent Words for Accurate Zero-Shot Composed Image Retrieval | ✓ Link | | 55.1 | | | | Context-I2W (CLIP L/14) | 2023-09-28 |
| Pretrain like Your Inference: Masked Tuning Improves Zero-Shot Composed Image Retrieval | ✓ Link | | 54.58 | | | | MTCIR (CLIP L/14) | 2023-11-13 |
| CompoDiff: Versatile Composed Image Retrieval With Latent Diffusion | ✓ Link | | 54.36 | | | | CompoDiff (CLIP L/14) | 2023-03-21 |
| iSEARLE: Improving Textual Inversion for Zero-Shot Composed Image Retrieval | ✓ Link | | 54.05 | | | | iSEARLE-XL-OTI (CLIP L/14) | 2024-05-05 |
| iSEARLE: Improving Textual Inversion for Zero-Shot Composed Image Retrieval | ✓ Link | | 54.00 | | | | iSEARLE-XL (CLIP L/14) | 2024-05-05 |
| Zero-Shot Composed Image Retrieval with Textual Inversion | ✓ Link | | 53.42 | | | | SEARLE | 2023-03-27 |
| Language-only Efficient Training of Zero-shot Composed Image Retrieval | ✓ Link | | 53.25 | | | | LinCIR (CLIP L/14) | 2023-12-04 |
| Zero-Shot Composed Image Retrieval with Textual Inversion | ✓ Link | | 52.48 | | | | SEARLE-XL | 2023-03-27 |
| Pic2Word: Mapping Pictures to Words for Zero-shot Composed Image Retrieval | ✓ Link | | 51.70 | | | | Pic2Word | 2023-02-06 |
| "This is my unicorn, Fluffy": Personalizing frozen vision-language representations | ✓ Link | | 43.49 | | | | PALAVRA | 2022-04-04 |