Research map / Vision-language and multimodal / Cross-modal learning
KODA: Contrastive Representation Comparison and Alignment for Vision-Language Foundation Models
Youqi WU, Mohammad Jalali, Farzan Farnia
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 |
|---|---|---|---|
| fine-tune CLIP on KODA-selected slices with kernel alignment to DINOv2 | CLIP · FFHQ KODA slices | AMI: 0.25 -> 0.78 | Improved |
| fine-tune CLIP on KODA-selected slices with kernel alignment to BLIP | CLIP · MSCOCO KODA slices | AMI: 0.52 -> 0.91 | Improved |
| fine-tune CLIP on KODA slices (ARI) | CLIP · FFHQ KODA slices | ARI: 0.19 -> 0.70 | Improved |
| evaluate on KODA-selected subset instead of full MSCOCO | CLIP vs SigLIP retrieval · MSCOCO retrieval | I2T R@1: CLIP 32.64 -> 18.00; SigLIP 42.64 -> 34.00 | Hurt |
| use joint image-text embeddings instead of image-only embeddings for KODA (ablation) | KODA · MSCOCO discrepancy directions | cluster coherence (qualitative) | Improved |
| add explicit weak-clusterability constraint under reference kernel (vs kernel-difference of SPEC) | KODA · contrastive embedding clustering | discrepancy discovery (qualitative) | Improved |
Closest papers by meaning
- Seeing the Image: Prioritizing Visual Correlation by Contrastive Alignment (NeurIPS 2024 · similarity 0.89)
- Improving fine-grained understanding in image-text pre-training (ICML 2024 · similarity 0.88)
- Post-pre-training for Modality Alignment in Vision-Language Foundation Models (CVPR 2025 · similarity 0.88)
- Scaling Language-centric Omnimodal Representation Learning (NeurIPS 2025 · similarity 0.88)
- Synthesize Diagnose and Optimize: Towards Fine-Grained Vision-Language Understanding (CVPR 2024 · similarity 0.88)
- DH-Set: Improving Vision-Language Alignment with Diverse and Hybrid Set-Embeddings Learning (CVPR 2025 · 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.
