Research map / Vision-language and multimodal / Multimodal LLMs
CalliReader: Contextualizing Chinese Calligraphy via an Embedding-Aligned Vision-Language Model
Yuxuan Luo, Jiaqi Tang, Chenyi Huang, Feiyang Hao, Zhouhui Lian
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 e-IT embedding instruction tuning (ablation) | CalliReader · CalliBench Hard tier | Macro-F1: 0.390 -> 0.609 | Improved |
| add e-IT embedding instruction tuning (ablation) | CalliReader · CalliBench Author VQA | Acc (%): 79.31 -> 95.06 | Improved |
| add character-wise slicing and CalliAlign to InternVL2-8B (ablation) | InternVL2-8B · CalliBench Hard tier | Macro-F1: 0.324 -> 0.390 | Improved |
| add character-wise slicing and CalliAlign to InternVL2-8B (ablation) | InternVL2-8B · CalliBench Easy tier | Macro-F1: 0.617 -> 0.779 | Improved |
| use four blocks instead of two in CalliAlign (ablation) | CalliAlign · label prediction | Acc: 0.865 -> 0.897 | Improved |
| use layer normalization instead of global normalization (ablation) | CalliAlign · label prediction | Acc: 0.791 -> 0.897 | Improved |
| add ratio loss Lrat (ablation) | CalliAlign · label prediction | Acc: 0.897 -> 0.695 | Hurt |
| add contrastive distillation loss LCRD (ablation) | CalliAlign · label prediction | Acc: 0.897 -> 0.813 | Hurt |
| add slicing and CalliAlign plug-ins (ablation) | CalliReader · Hard tier latency | s/page: 41.29 -> 33.50 | Improved |
| add slicing and CalliAlign plug-ins (ablation) | CalliReader · Easy tier latency | s/page: 1.32 -> 1.48 | Hurt |
| single-task LoRA fine-tuning on page recognition | Qwen2.5-VL-7B · Contextual VQA Intent | Acc (%): 70.80 -> 14.43 | Hurt |
| single-task LoRA fine-tuning on page recognition | InternVL2-8B · Contextual VQA Layout | Acc (%): 45.66 -> 39.70 | No clear effect |
Closest papers by meaning
- DeepCalliFont: Few-Shot Chinese Calligraphy Font Synthesis by Integrating Dual-Modality Generative Models (AAAI 2024 · similarity 0.87)
- MosaicDoc: A Large-Scale Bilingual Benchmark for Visually Rich Document Understanding (AAAI 2026 · similarity 0.86)
- ASCIIEval: Benchmarking Models' Visual Perception in Text Strings via ASCII Art (ICLR 2026 · similarity 0.86)
- Towards Automated Chinese Ancient Character Restoration: A Diffusion-Based Method with a New Dataset (AAAI 2024 · similarity 0.85)
- Decoupling Layout from Glyph in Online Chinese Handwriting Generation (ICLR 2025 · similarity 0.85)
- A Multi-Style Chinese Characters Writing Intelligent Tool Based on Small-scale Training Data (AAAI 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.
