Research map / Vision-language and multimodal / Multimodal LLMs
HSD: Training-Free Acceleration for Document Parsing Vision-Language Models with Hierarchical Speculative Decoding
Wenhui Liao, Hongliang Li, Pengyu Xie, Xinyu Cai, Yufan Shen, Yi Xin, Qi Qin, Shenglong Ye, Tianbin Li, Ming Hu, Junjun He, Yihao Liu and 6 more
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 Stage 2 page-level verification after Stage 1 (ablation) | HSD with dots.ocr · OmniDocBench v1.5 | accuracy: 70.47 -> 88.81 | Improved |
| add hierarchical two-stage design to pipeline drafter (ablation) | HSD with dots.ocr · OmniDocBench v1.5 | end-to-end speedup: 2.09x -> 2.42x | Improved |
| add hierarchical two-stage design to pipeline drafter (ablation) | HSD with dots.ocr · OmniDocBench v1.5 | AAL: 2.49 -> 3.98 | Improved |
| use a fast pipeline drafter instead of autoregressive decoding (ablation) | dots.ocr · OmniDocBench v1.5 | end-to-end speedup: 1.00x -> 2.09x | Improved |
| add HSD to end-to-end parser | HunyuanOCR · OmniDocBench v1.5 | mean latency (s/sample): 30.47 -> 10.96 | Improved |
| add HSD to end-to-end parser | HunyuanOCR · OmniDocBench v1.5 | accuracy: 94.10 -> 94.02 | No clear effect |
| add HSD to end-to-end parser | dots.ocr · OmniDocBench v1.5 | accuracy: 86.73 -> 88.81 | No clear effect |
| add HSD on top of visual token compression | DeepSeek-OCR · OmniDocBench v1.5 | end-to-end speedup: 1.00x -> 1.56x | Improved |
| use HSD instead of ViSpec speculative decoding | ViSpec · OmniDocBench v1.5 | end-to-end speedup: 1.75x -> 2.12x | Improved |
| lower acceptance threshold tau below 0.75 (ablation) | HSD with dots.ocr · OmniDocBench v1.5 | parsing accuracy | Hurt |
| lower acceptance threshold tau (ablation) | HSD with dots.ocr · OmniDocBench v1.5 | speedup | Improved |
| degrade pipeline draft quality with injected noise (ablation) | HSD with dots.ocr · OmniDocBench v1.5 | end-to-end speedup | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ViSpec — ViSpec: Accelerating Vision-Language Models with Vision-Aware Speculative Decoding (NeurIPS 2025) · better on 3 of 3 shared results
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
- Towards Real-World Document Parsing via Realistic Scene Synthesis and Document-Aware Training (CVPR 2026 · similarity 0.89)
- Boosting Document Parsing Efficiency and Performance with Coarse-to-Fine Visual Processing (CVPR 2026 · similarity 0.87)
- MPDocBench-Parse: Benchmarking Practical Multi-page Document Parsing (NeurIPS 2026 · similarity 0.87)
- OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations (CVPR 2025 · similarity 0.87)
- RAPID: Long-Context Inference with Retrieval-Augmented Speculative Decoding (ICML 2025 · similarity 0.86)
- ViSpec: Accelerating Vision-Language Models with Vision-Aware Speculative Decoding (NeurIPS 2025 · similarity 0.85)
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.
