Research map / Vision-language and multimodal / Cross-modal learning
SOLAR: Self-supervised Joint Learning for Symmetric Multimodal Retrieval
Wenjie Yang, Hang Yu, Yuyu Guo, Peng Di
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 disentanglement (masked ITC with GLA, GD, LD) instead of plain contrastive alignment only (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 81.5 -> 86.5 | Improved |
| add masked ITC loss (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 82.6 -> 86.5 | Improved |
| add global-to-local alignment loss (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 80.8 -> 86.5 | Improved |
| add global unimodal distillation (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 84.9 -> 86.5 | Improved |
| add local unimodal distillation (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 82.4 -> 86.5 | Improved |
| add evolutionary mask schedule (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 86.0 -> 86.5 | Improved |
| use annealed evolutionary mask instead of static hard mask (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 84.8 -> 86.5 | Improved |
| add projection heads (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 82.6 -> 86.5 | Improved |
| add Stage 2 self-supervised contrastive training after Stage 1 (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 82.2 -> 86.5 | Improved |
| add constructed negatives (masking the difference set) (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 84.0 -> 86.5 | Improved |
| mask the semantic intersection for positives instead of random masking (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 83.3 -> 86.5 | Improved |
| use semantic segmentation masking instead of patch-level masking (ablation) | SOLAR-B+D · sym-MM2MM | Avg: 81.5 -> 86.5 | 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.
- LamRA — LamRA: Large Multimodal Model as Your Advanced Retrieval Assistant (CVPR 2025) · better on 2 of 2 shared results
- MM-Embed — MM-EMBED: UNIVERSAL MULTIMODAL RETRIEVAL WITH MULTIMODAL LLMS (ICLR 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
- MetaEmbed: Scaling Multimodal Retrieval at Test-Time with Flexible Late Interaction (ICLR 2026 · similarity 0.88)
- MM-EMBED: UNIVERSAL MULTIMODAL RETRIEVAL WITH MULTIMODAL LLMS (ICLR 2025 · similarity 0.88)
- U-MARVEL: Unveiling Key Factors for Universal Multimodal Retrieval via Embedding Learning with MLLMs (ICLR 2026 · similarity 0.88)
- Your Embedding Model Is SMARTer Than You Think (NeurIPS 2026 · similarity 0.88)
- RMIR: A Benchmark Dataset for Reasoning-Intensive Multimodal Image Retrieval (CVPR 2026 · similarity 0.87)
- MRMR: A Realistic and Expert-Level Multidisciplinary Benchmark for Reasoning-Intensive Multimodal Retrieval (ICLR 2026 · similarity 0.87)
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.
