Research map / Vision-language and multimodal / Human and embodied modeling
Learning Sample-wise Rank-Aware Interpolation Weights for Composed Visual Data Retrieval
Boseung Jeong, Taegyu Park, Donghyeon Kwon, Hyunsouk Cho, Suha Kwak
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add sample-wise predicted weight (stage 2) on frozen BLIP-2 (ablation) | Average baseline · WebVid-CoVR-Test (ablation) | R@1: 45.66 -> 48.08 | Improved |
| use two-stage training instead of end-to-end training (ablation) | SRAIN · WebVid-CoVR-Test (ablation) | R@1: 57.36 -> 61.07 | Improved |
| add weight predictor over fixed alpha=0.5 (Stage 1 only) (ablation) | SRAIN · WebVid-CoVR-Test (ablation) | R@1: 60.50 -> 61.07 | Improved |
| add stage 1 encoder fine-tuning to stage 2 predictor (ablation) | SRAIN · WebVid-CoVR-Test (ablation) | R@1: 48.08 -> 61.07 | Improved |
| add conditioning tokens C_i from hard negatives (ablation) | SRAIN · WebVid-CoVR-Test (ablation) | R@1: 60.33 -> 61.07 | Improved |
| add memory bank at test time (ablation) | SRAIN · WebVid-CoVR-Test (ablation) | R@1: 60.62 -> 61.07 | Improved |
| add memory bank conditioning in zero-shot CIRCO (ablation) | SRAIN · CIRCO (zero-shot) | mAP@5: 23.79 -> 26.16 | Improved |
| use memory prototypes instead of random vectors as conditioning (ablation) | SRAIN · CIRCO (zero-shot) | mAP@5: 23.72 -> 26.16 | Improved |
| use oracle optimal weights instead of predicted weights (ablation) | SRAIN · WebVid-CoVR-Test (ablation) | R@1: 61.07 -> 71.95 | Improved |
| use SRAIN over best fixed interpolation baseline | Average (alpha=0.5) · WebVid-CoVR-Test | R@1: 58.69 -> 61.07 | Improved |
| use SRAIN instead of CoVR-2 cross-attention | CoVR-2 · WebVid-CoVR-Test | R@1: 59.82 -> 61.07 | Improved |
| use SRAIN instead of CoVR with BLIP backbone | CoVR · WebVid-CoVR-Test (BLIP backbone) | R@1: 55.95 -> 56.91 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CoVR — COVR: Collaborative Optimization of VLMs and RL Agent for Visual-Based Control (AAAI 2026) · better on 8 of 8 shared results
- BASIC — BASIC: Boosting Visual Alignment with Intrinsic Refined Embeddings in Multimodal Large Language Models (ICCV 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Thinking Fast and Slow: Efficient Text-to-Visual Retrieval With Transformers (CVPR 2021 · similarity 0.88)
- Instance-Level Composed Image Retrieval (NeurIPS 2025 · similarity 0.87)
- ReCALL: Recalibrating Capability Degradation for MLLM-based Composed Image Retrieval (CVPR 2026 · similarity 0.87)
- NUDGE: Lightweight Non-Parametric Fine-Tuning of Embeddings for Retrieval (ICLR 2025 · similarity 0.86)
- DaWin: Training-free Dynamic Weight Interpolation for Robust Adaptation (ICLR 2025 · similarity 0.86)
- QuARI: Query Adaptive Retrieval Improvement (NeurIPS 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
