Research map / Vision-language and multimodal / Human and embodied modeling
UniRec-0.1B: Unified Text and Formula Recognition with 0.1B Parameters
Yongkun Du, Zhineng Chen, Yazhen Xie, Weikang Bai, Hao Feng, Wei Shi, Yuchen Su, Can Huang, Yu-Gang Jiang
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 hierarchical supervision training (line and paragraph tokens) (ablation) | UniRec-0.1B · UniRec-Bench average | edit distance: 0.113 -> 0.100 | Improved |
| add hierarchical supervision training (ablation) | UniRec-0.1B · UniRec-Bench text | edit distance: 0.050 -> 0.038 | Improved |
| add hierarchical supervision training (ablation) | UniRec-0.1B · UniRec-Bench formula | edit distance: 0.144 -> 0.134 | Improved |
| add hierarchical supervision training (ablation) | UniRec-0.1B · UniRec-Bench mix | edit distance: 0.143 -> 0.128 | Improved |
| add hierarchical supervision training (ablation) | UniRec-0.1B · UniRec-Bench PPT2PDF | edit distance: 0.091 -> 0.063 | Improved |
| add semantics-decoupled tokenizer on top of HST-less model (ablation) | UniRec-0.1B w/o HST · UniRec-Bench formula | edit distance: 0.255 -> 0.144 | Improved |
| add semantics-decoupled tokenizer on top of HST-less model (ablation) | UniRec-0.1B w/o HST · UniRec-Bench text | edit distance: 0.062 -> 0.050 | Improved |
| add semantics-decoupled tokenizer on top of HST-less model (ablation) | UniRec-0.1B w/o HST · UniRec-Bench mix | edit distance: 0.161 -> 0.143 | Improved |
| add both HST and SDT (ablation) | UniRec-0.1B · UniRec-Bench average | edit distance: 0.159 -> 0.100 | Improved |
| replace MinerU2.5 text/formula recognizer with UniRec-0.1B | MinerU2.5 · OmniDocBench overall | edit distance: 0.143 -> 0.120 | Improved |
| replace MinerU2.5 text/formula recognizer with UniRec-0.1B | MinerU2.5 · page parsing time | seconds: 42.72 -> 6.2 | Improved |
| replace PaddleOCR-VL text/formula recognizer with UniRec-0.1B | PaddleOCR-VL · OmniDocBench overall | edit distance: 0.115 -> 0.113 | Improved |
Closest papers by meaning
- Image Over Text: Transforming Formula Recognition Evaluation with Character Detection Matching (CVPR 2025 · similarity 0.86)
- FLAME: A Small Language Model for Spreadsheet Formulas (AAAI 2024 · similarity 0.84)
- MPDocBench-Parse: Benchmarking Practical Multi-page Document Parsing (NeurIPS 2026 · similarity 0.84)
- DocLayLLM: An Efficient Multi-modal Extension of Large Language Models for Text-rich Document Understanding (CVPR 2025 · similarity 0.83)
- OpenWebMath: An Open Dataset of High-Quality Mathematical Web Text (ICLR 2024 · similarity 0.83)
- OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations (CVPR 2025 · similarity 0.83)
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.
