AI Paper Atlas

Research map / Large language models / Agents, code and math

MCHDoc: A Comprehensive Benchmark for Reading Multi-Carrier Chinese Historical Documents

Yijun Sheng, Shipeng Zhu, Ruijia Zuo, Na Nie, Hui Xue

CVPR 2026 · Large language models › Agents, code and math · approach: Mathematical and structural reasoning benchmarks

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 changedWhereMeasuredEffect
fine-tune Qwen2.5-VL-3B on multi-carrier historical data (SFT) (ablation)Qwen2.5-VL-3B · MCHDoc page-level, AncientBook1-NED: 29.55 -> 52.53Improved
fine-tune Qwen2.5-VL-3B (SFT) (ablation)Qwen2.5-VL-3B · MCHDoc page-level, Inscription1-NED: 48.96 -> 48.44No clear effect
LoRA fine-tune InternVL3.5-1B (ablation)InternVL3.5-1B · MCHDoc character-level, AncientBookaccuracy: 13.6 -> 62.6Improved
LoRA fine-tune InternVL3.5-1B (ablation)InternVL3.5-1B · MCHDoc character-level, JianDuaccuracy: 0.3 -> 17.6Improved
add LLM post-correction with internal knowledge onlyDoubao-Seed-1.6-VL output · MCHDoc, Inscription1-NED: 65.90 -> 55.81Hurt
add LLM post-correction with internal knowledge onlyDoubao-Seed-1.6-VL output · MCHDoc, AncientBook1-NED: 56.52 -> 51.86Hurt
add external knowledge base retrieval to post-correctionGemini-2.5-Flash-Think · MCHDoc, AncientBook1-NED: 51.86 -> 65.64Improved
add external knowledge base retrieval to post-correctionDeepseek-V3.1-Think · MCHDoc, Calligraphy1-NED: 43.20 -> 72.92Improved
add external knowledge base retrieval to post-correctionGemini-2.5-Flash-Think · MCHDoc, Inscription1-NED: 55.81 -> 42.51Hurt
add external retrieval to post-correction on JianDuDoubao-Seed-1.6-VL output · MCHDoc, JianDu1-NED: 18.82 -> 17.91 (best, Deepseek-V3.1-Think)No clear effect
use GPT series for long historical pagesGPT-5 · MCHDoc page-level, AncientBook1-NED: 56.52 -> 6.87 (vs Doubao)Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy