Research map / Vision-language and multimodal / CLIP and zero-shot
Fix-CLIP: Dual-Branch Hierarchical Contrastive Learning via Synthetic Captions for Better Understanding of Long Text
Bingchao Wang, Zhiwei Ning, Jianyu Ding, Xuanang Gao, Yin Li, Dongsheng Jiang, Jie Yang, Wei Liu
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 dual-branch training pipeline (ablation) | Long-CLIP baseline · DCI (ViT-L/14) | I2T R@1: 51.7 -> 60.5 | Improved |
| add dual-branch training pipeline (ablation) | Long-CLIP baseline · DCI (ViT-L/14) | T2I R@1: 67.4 -> 61.4 | Hurt |
| add hierarchical feature alignment (ablation) | Long-CLIP baseline · DCI (ViT-L/14) | I2T R@1: 51.7 -> 53.3 | Improved |
| add regional prompts (ablation) | dual-branch model · DCI (ViT-L/14) | I2T R@1: 60.5 -> 62.4 | Improved |
| add unidirectional mask (ablation) | dual-branch + regional prompts · DCI (ViT-L/14) | I2T R@1: 62.4 -> 63.5 | Improved |
| combine all four components (ablation) | Fix-CLIP · DCI (ViT-L/14) | T2I R@1: 63.1 -> 66.7 | Improved |
| use separate position embedding for short and long text (ablation) | Fix-CLIP · DCI (ViT-L/14) | I2T R@1: 62.6 -> 64.3 | Improved |
| preserve masked patches instead of discarding them (ablation) | Fix-CLIP · DCI (ViT-L/14) | I2T R@1: 62.7 -> 65.1 | Improved |
| use 4 regional prompts instead of 8 (ablation) | Fix-CLIP · DCI (ViT-L/14) | T2I R@1: 65.4 -> 66.7 | Improved |
| use 1 regional prompt instead of 4 (ablation) | Fix-CLIP · COCO I2T (ViT-L/14) | R@1: 63.4 -> 63.6 | Mixed |
| apply hierarchical contrastive loss from group 4 to 6 instead of none (ablation) | Fix-CLIP · DCI (ViT-L/14) | T2I R@1: 63.1 -> 66.7 | Improved |
| scale synthetic training data from 1M to 30M | Fix-CLIP ViT-B/16 · Long-text retrieval avg | R@1: 82.6 -> 89.8 | Improved |
Later papers that compared with this method
1 later accepted paper reports FIX-CLIP 1M as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Role-SynthCLIP: A Role-Play Driven Diverse Synthetic Data Approach (CVPR 2026 · better on 2 of 2 shared results)
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.
- Long-CLIP (1M) — Long-CLIP: Unlocking the Long-Text Capability of CLIP (ECCV 2024) · better on 10 of 10 shared results
- LoTLIP (100M) — LoTLIP: Improving Language-Image Pre-training for Long Text Understanding (NeurIPS 2024) · 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
- Retaining Knowledge and Enhancing Long-Text Representations in CLIP through Dual-Teacher Distillation (CVPR 2025 · similarity 0.90)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 2025 · similarity 0.89)
- FineLIP: Extending CLIP's Reach via Fine-Grained Alignment with Longer Text Inputs (CVPR 2025 · similarity 0.89)
- GOAL: Global-local Object Alignment Learning (CVPR 2025 · similarity 0.88)
- Mining Fine-Grained Image-Text Alignment for Zero-Shot Captioning via Text-Only Training (AAAI 2024 · similarity 0.88)
- Enhancing Compositional Reasoning in CLIP via Reconstruction and Alignment of Text Descriptions (NeurIPS 2025 · similarity 0.88)
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.
