Research map / Vision-language and multimodal / Cross-modal learning
FLARE: Fully Integration of Vision-Language Representations for Deep Cross-Modal Understanding
Zheng Liu, Mengjie Liu, Jingzhou Chen, Jingwei Xu, Bin CUI, Conghui He, Wentao Zhang
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 Text-Guided Vision Encoding (ablation) | baseline VLM · MMBench-EN | accuracy: 72.5 -> 73.7 | Improved |
| add Context-Aware Alignment Decoding (ablation) | baseline VLM · TextVQA | accuracy: 61.5 -> 64.4 | Improved |
| add Dual-Semantic Mapping Loss on top of text-guided encoding (ablation) | FLARE-L · MMBench-EN | accuracy: 73.7 -> 74.5 | Improved |
| add Context-Aware Decoding to A+B (ablation) | FLARE-L · MMStar | accuracy: 44.1 -> 45.5 | Mixed |
| add all three components vs baseline (ablation) | FLARE-L · TextVQA | accuracy: 61.5 -> 65.6 | Improved |
| add synthetic text-driven VQA data | FLARE-L 8B · MMBench-EN | accuracy: 79.4 -> 81.8 | Improved |
| add synthetic data | FLARE-L 8B · MMStar | accuracy: 54.6 -> 54.5 | No clear effect |
| add synthetic data | FLARE-X 8B · ChartQA | accuracy: 80.3 -> 83.4 | Improved |
| use dynamic resolution (FLARE-X) over fixed (FLARE-L) | FLARE 8B · TextVQA | accuracy: 76.2 -> 79.7 | Improved |
| insert more semantic exchange layers per insertion point (6 interval, 3 vs 1) (ablation) | FLARE (Vicuna-7B+CLIP) · MMBench-EN | accuracy: 70.4 -> 68.1 | Hurt |
| FLARE-L 3B vs Florence-VL/Cambrian-1 8B | FLARE-L 3B · 16 benchmarks | accuracy | Improved |
| FLARE vs Qwen2.5VL on OCRBench (same backbone, 7B) | FLARE 7B · OCRBench | accuracy: 88.5 -> 78.4 | Hurt |
Closest papers by meaning
- From Pixels to Words -- Towards Native Vision-Language Primitives at Scale (ICLR 2026 · similarity 0.88)
- LoMo: Local Modality Substitution for Deeper Vision-Language Fusion (NeurIPS 2026 · similarity 0.88)
- Modality-Specialized Synergizers for Interleaved Vision-Language Generalists (ICLR 2025 · similarity 0.88)
- Analyzing Fine-Grained Alignment and Enhancing Vision Understanding in Multimodal Language Models (NeurIPS 2025 · similarity 0.88)
- Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in Vision-Language Alignment (ICLR 2025 · similarity 0.87)
- Deep Pre-Alignment for VLMs (ICML 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.
