Research map / Vision-language and multimodal / Human and embodied modeling
Modeling Relationships in Referential Expressions With Compositional Modular Networks
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 pairwise relationship modeling to region localization | CMN · Synthetic shape dataset | Accuracy: 46.27% -> 99.99% | Helped |
| Add pairwise relationship modeling to region localization | CMN · Visual Genome, subject retrieval | P@1-subj: 41.20% -> 43.81% | Helped |
| Add pairwise relationship modeling to region localization | CMN · Visual Genome, subject-object pair retrieval | P@1-pair: 23.37% -> 26.56% | Helped |
| Use subject-object-GT rather than subject-GT supervision | CMN · Visual Genome, subject retrieval | P@1-subj: 43.81% -> 44.24% | Helped |
| Use subject-object-GT rather than subject-GT supervision | CMN · Visual Genome, subject-object pair retrieval | P@1-pair: 26.56% -> 28.52% | Helped |
| Add inter-object relationship modeling | CMN · Google-Ref | P@1: 66.5% -> 69.3% | Helped |
| Replace learned expression parsing with an external parser (ablation) | CMN · Google-Ref | P@1: 69.3% -> 53.5% | Hurt |
| Replace learned expression parsing with an external parser (ablation) | CMN · Visual-7W pointing questions | Accuracy: 72.53% -> 61.66% | Hurt |
| Add inter-object relationship modeling | CMN · Visual-7W pointing questions | Accuracy: 71.61% -> 72.53% | Helped |
| Add visual features for both regions to the relationship module (ablation) | CMN relationship module · Relationship module training | Performance | No clear effect |
Closest papers by meaning
- Cross-Modal Relationship Inference for Grounding Referring Expressions (CVPR 2019 · similarity 0.90)
- Causal Graphical Models for Vision-Language Compositional Understanding (ICLR 2025 · similarity 0.85)
- Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation (CVPR 2024 · similarity 0.85)
- Zero-shot Referring Expression Comprehension via Structural Similarity Between Images and Captions (CVPR 2024 · similarity 0.84)
- ClawMachine: Learning to Fetch Visual Tokens for Referential Comprehension (ICLR 2025 · similarity 0.84)
- DViN: Dynamic Visual Routing Network for Weakly Supervised Referring Expression Comprehension (CVPR 2025 · similarity 0.84)
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.