[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-77538":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":13,"stars7d":16,"stars30d":17,"stars90d":15,"forks30d":15,"starsTrendScore":18,"compositeScore":19,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":20,"fork":20,"defaultBranch":21,"hasWiki":22,"hasPages":20,"topics":23,"createdAt":9,"pushedAt":9,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":15,"starSnapshotCount":15,"syncStatus":14,"lastSyncTime":27,"discoverSource":28},77538,"Chronicles-OCR","VirtualLUOUCAS\u002FChronicles-OCR","VirtualLUOUCAS","Chronicles-OCR: A Cross-Temporal Perception Benchmark for the Evolutionary Trajectory of Chinese Characters (Seven Chinese Scripts, 2800 images)",null,"Python",237,12,1,2,0,3,213,5,53.84,false,"main",true,[],"2026-06-12 04:01:21","# Chronicles-OCR\n\n**A Cross-Temporal Perception Benchmark for the Evolutionary Trajectory of Chinese Characters**\n\n\u003Cp align=\"center\">\n  \u003Ca href=\"README_ZH.md\">中文版\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.11960\">Paper\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FVirtualLUOUCAS\u002FChronicles-OCR\">GitHub\u003C\u002Fa> •\n  \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002FVirtualLUO\u002FChronicles-OCR\">HuggingFace\u003C\u002Fa>\n\u003C\u002Fp>\n\n## Overview\n\n**Chronicles-OCR** is the first comprehensive benchmark specifically designed to evaluate the cross-temporal visual perception capabilities of VLLMs across the complete evolutionary trajectory of Chinese characters — the **\"Seven Chinese Scripts\"**.\n\nCurated in collaboration with top-tier institutional domain experts (the Key Laboratory of Oracle Bone Inscription Information Processing at Anyang Normal University and the Palace Museum), the dataset comprises **2,800 strictly balanced images** encompassing highly diverse physical media, ranging from tortoise shells to paper-based calligraphy.\n\n\u003Cp align=\"center\">\n  \u003Cimg src=\"assets\u002Foverview.png\" width=\"95%\" alt=\"Chronicles-OCR Overview\">\n\u003C\u002Fp>\n\n## The Seven Chinese Scripts\n\nThe **\"Seven Chinese Scripts\" (汉字七体)** refer to the seven canonical script forms that emerged throughout the evolution of Chinese characters over more than 5,000 years:\n\n1. **Oracle Bone Script (甲骨文)** — The earliest known mature Chinese writing system, carved on tortoise shells and animal bones during the Shang Dynasty. Characters feature strong pictographic qualities with thin, angular strokes and unstandardized layouts.\n2. **Bronze Script (金文)** — Cast on ceremonial bronze vessels during the Shang and Zhou Dynasties. Strokes are thicker and rounder, with progressively more regularized and aesthetic structures.\n3. **Seal Script (篆书)** — Standardized after the Qin unification of China. Features pronounced curvilinear symmetry and fixed structural patterns, marking the transition from regional variants to a unified writing system.\n4. **Clerical Script (隶书)** — Emerged during the Qin–Han transition, flattening characters and replacing curves with angular strokes. Represents a critical turning point — the watershed between ancient and modern Chinese characters.\n5. **Regular Script (楷书)** — Established in the late Han and Wei-Jin periods with strict square structures and standardized strokes. Remains the dominant formal script to this day.\n6. **Cursive Script (草书)** — Developed for rapid, informal writing. Uses continuous, connected strokes that often eliminate independent character boundaries, ranging from the restrained Zhang Cao to the unbounded Kuang Cao.\n7. **Running Script (行书)** — A fluid yet legible intermediate style between Regular and Cursive scripts, widely used from the Eastern Han Dynasty onward. Wang Xizhi's *Preface to the Orchid Pavilion* is its most celebrated exemplar.\n\nAmong these, the first five (Oracle Bone → Regular) successively served as formal writing systems in their respective eras, while Cursive and Running scripts developed primarily as auxiliary styles for informal and rapid writing.\n\n## Benchmark Statistics\n\n| Item | Details |\n|------|---------|\n| Total Images | 2,800 (400 per script × 7 scripts) |\n| Script Coverage | All Seven Chinese Scripts |\n| Annotation | Stage-Adaptive: character-level for archaic, paragraph-level for mature scripts |\n| Expert Partners | Anyang Normal University (Oracle Bone), Palace Museum (Clerical–Cursive) |\n| Tasks | 4 evaluation tasks |\n\n## Evaluation Tasks\n\n| Task | Short Name | Scope | Metric |\n|------|-----------|-------|--------|\n| Cross-period Character Spotting | Spotting | Oracle Bone, Bronze, Seal | F1 @ IoU > 0.75 |\n| Fine-grained Archaic Character Recognition | Recognition | Oracle Bone, Bronze, Seal | Exact-match Accuracy |\n| Ancient Text Parsing | Parsing | All Seven Scripts | 1 − NED (Levenshtein) |\n| Script Classification | Classification | All Seven Scripts | Accuracy |\n\n## 🏆 Leaderboard\n\n### Archaic Scripts (Oracle Bone, Bronze, Seal)\n\n| Model | Think | Avg Spot. | Avg Fine. | Avg Pars. | Avg Class. | OB Spot. | OB Fine. | OB Pars. | OB Class. | Br Spot. | Br Fine. | Br Pars. | Br Class. | Se Spot. | Se Fine. | Se Pars. | Se Class. |\n|:------|:-----:|:---------:|:---------:|:---------:|:----------:|:--------:|:--------:|:--------:|:---------:|:--------:|:--------:|:--------:|:---------:|:--------:|:--------:|:--------:|:---------:|\n| **Open-Source Models** | | | | | | | | | | | | | | | | | |\n| InternVL3.5-8B | | 0.1 | 6.0 | 0.07 | 56.7 | 0.0 | 1.1 | 0.01 | 86.2 | 0.0 | 2.2 | 0.03 | 7.0 | 0.2 | 14.5 | 0.17 | 77.0 |\n| InternVL3.5-A28B | | 0.5 | 15.7 | 0.13 | 79.0 | 0.0 | 2.5 | 0.02 | 96.3 | 0.4 | 7.8 | 0.08 | 79.2 | 1.0 | 36.8 | 0.29 | 61.5 |\n| Qwen2.5-VL-7B | | 0.0 | 7.4 | 0.07 | 71.8 | 0.0 | 4.0 | 0.02 | 93.8 | 0.0 | 4.5 | 0.04 | 22.5 | 0.0 | 13.8 | 0.14 | 99.2 |\n| Qwen2.5-VL-72B | | 0.0 | 0.0 | 0.07 | 74.2 | 0.0 | 0.0 | 0.01 | 98.0 | 0.0 | 0.0 | 0.04 | 26.0 | 0.0 | 0.0 | 0.16 | 98.5 |\n| Qwen3-VL-2B | | 2.1 | 10.7 | 0.12 | 73.0 | 0.0 | 1.4 | 0.00 | 96.6 | 0.8 | 6.8 | 0.06 | 36.5 | 5.7 | 24.0 | 0.31 | 85.8 |\n| Qwen3-VL-8B | | 3.4 | 17.3 | 0.18 | 73.7 | 0.2 | 3.4 | 0.01 | 98.6 | 2.5 | 11.0 | 0.10 | 24.0 | 7.5 | 37.5 | 0.42 | 98.5 |\n| Qwen3-VL-8B | ✓ | 1.0 | 9.1 | 0.09 | 67.3 | 0.0 | 3.7 | 0.03 | 97.7 | 0.2 | 7.0 | 0.05 | 31.8 | 2.8 | 16.8 | 0.20 | 72.5 |\n| Qwen3-VL-A22B | | 7.8 | 17.5 | 0.19 | 91.8 | 0.3 | 5.4 | 0.01 | 99.2 | 6.5 | 12.2 | 0.12 | 80.2 | 16.6 | 35.0 | 0.43 | 96.0 |\n| Qwen3-VL-A22B | ✓ | 2.1 | 13.6 | 0.17 | 87.3 | 0.1 | 4.2 | 0.03 | 98.0 | 0.9 | 10.2 | 0.11 | 66.8 | 5.3 | 26.2 | 0.37 | 97.2 |\n| Qwen3.5-A3B | | 5.6 | 16.2 | 0.20 | 76.5 | 0.2 | 5.1 | 0.02 | 99.7 | 5.3 | 11.5 | 0.12 | 30.0 | 11.2 | 32.0 | 0.45 | **99.8** |\n| Qwen3.5-A17B | | 9.7 | 22.6 | 0.22 | 88.3 | 0.5 | 9.1 | 0.02 | 99.7 | 9.2 | 17.5 | 0.13 | 67.2 | 19.4 | 41.3 | 0.50 | 98.0 |\n| Gemma 4 31B it | | 2.3 | 7.0 | 0.04 | 70.0 | 0.0 | 3.1 | 0.01 | 72.6 | 1.0 | 6.5 | 0.03 | 74.8 | 6.0 | 11.2 | 0.10 | 62.7 |\n| MiniCPM-V 4.5 | ✓ | 0.0 | 4.8 | 0.02 | 73.8 | 0.0 | 2.5 | 0.01 | 95.2 | 0.0 | 5.5 | 0.03 | 18.0 | 0.1 | 9.0 | 0.04 | 82.5 |\n| Molmo 7B-D 0924 | | 0.0 | 0.1 | 0.00 | 24.2 | 0.0 | 0.0 | 0.01 | 40.8 | 0.0 | 0.2 | 0.00 | 0.0 | 0.0 | 0.0 | 0.00 | 20.5 |\n| Molmo 72B 0924 | | 0.0 | 0.3 | 0.00 | 34.7 | 0.0 | 0.5 | 0.00 | 28.0 | 0.0 | 0.5 | 0.00 | 0.8 | 0.0 | 0.0 | 0.00 | 82.0 |\n| Ovis2.6-30B-A3B | ✓ | 1.9 | 9.0 | 0.09 | 68.3 | 0.1 | 2.0 | 0.01 | 89.8 | 0.7 | 7.5 | 0.06 | 13.5 | 6.8 | 24.5 | 0.25 | 79.0 |\n| GLM-4.5V 108B | ✓ | 1.4 | 6.1 | 0.05 | 76.8 | 0.1 | 4.2 | 0.03 | **100** | 2.0 | 6.5 | 0.05 | 15.5 | 3.3 | 9.2 | 0.10 | 91.5 |\n| Kimi K2.5 | | 5.0 | **27.1** | **0.22** | 96.4 | 0.1 | 11.5 | 0.05 | **100** | 7.5 | 25.8 | 0.19 | 90.0 | 12.5 | **58.5** | **0.60** | 95.5 |\n| Kimi K2.5 | ✓ | 1.8 | 20.3 | 0.22 | 94.7 | 0.0 | 10.2 | **0.05** | 99.8 | 1.2 | 17.5 | 0.20 | 85.8 | 6.0 | 44.8 | 0.57 | 93.5 |\n| **Proprietary Models** | | | | | | | | | | | | | | | | | |\n| GPT-4o | | 0.1 | 1.5 | 0.02 | 82.0 | 0.0 | 0.5 | 0.01 | 96.5 | 0.0 | 1.0 | 0.02 | 46.8 | 0.3 | 4.5 | 0.06 | 89.0 |\n| GPT-5 | | 0.4 | 3.7 | 0.04 | 88.1 | 0.0 | 4.0 | 0.00 | 98.2 | 0.0 | 4.0 | 0.04 | 60.5 | 1.6 | 4.5 | 0.12 | 97.5 |\n| Seed 1.8 | | 9.2 | 20.6 | 0.16 | 94.7 | 0.4 | 9.2 | 0.03 | 99.5 | 9.4 | 15.8 | 0.17 | 80.5 | 26.7 | 45.0 | 0.42 | 99.0 |\n| Seed 1.8 | ✓ | 7.4 | 17.1 | 0.17 | **96.7** | 0.4 | 8.8 | 0.04 | 99.5 | 5.8 | 14.8 | 0.18 | 90.0 | 23.3 | 36.2 | 0.43 | 97.5 |\n| Seed 2.0 Pro | | **16.5** | 24.5 | 0.18 | 95.9 | **3.0** | 11.0 | 0.03 | 99.5 | **19.9** | **30.8** | 0.22 | **92.2** | **40.7** | 41.5 | 0.43 | 93.8 |\n| Seed 2.0 Pro | ✓ | 15.3 | 23.3 | 0.21 | 96.6 | 2.4 | 11.2 | 0.04 | 99.8 | 17.8 | 26.0 | **0.26** | **92.2** | 39.1 | 37.5 | 0.49 | 94.5 |\n| MiMo-V2-Omni | ✓ | 0.4 | 8.6 | 0.08 | 87.7 | 0.0 | 6.5 | 0.04 | 99.5 | 0.2 | 8.0 | 0.07 | 58.5 | 1.5 | 9.8 | 0.15 | 93.0 |\n| Gemini 2.5 Pro | ✓ | 0.8 | 7.5 | 0.07 | 87.5 | 0.0 | 5.8 | 0.04 | 99.5 | 0.2 | 7.0 | 0.06 | 80.5 | 2.8 | 10.8 | 0.14 | 70.2 |\n| Gemini 3.1 Pro | ✓ | 2.6 | 19.5 | 0.15 | 93.8 | 0.0 | **14.0** | 0.05 | 99.5 | 2.5 | 22.5 | 0.18 | 84.5 | 7.8 | 32.2 | 0.32 | 93.2 |\n| Claude Opus 4.7 | ✓ | 0.4 | 10.0 | 0.08 | 90.4 | 0.0 | 4.8 | 0.03 | 93.8 | 0.1 | 9.5 | 0.05 | 80.5 | 1.4 | 21.5 | 0.21 | 93.8 |\n\n> **OB** = Oracle Bone, **Br** = Bronze, **Se** = Seal. **Bold** = best, scores are H-mean (Spot.), Accuracy (Fine.\u002FClass.), NED (Pars.).\n\n### Mature Scripts (Clerical, Regular, Running, Cursive)\n\n| Model | Think | Avg Pars. | Avg Class. | Cl Pars. | Cl Class. | Re Pars. | Re Class. | Ru Pars. | Ru Class. | Cu Pars. | Cu Class. |\n|:------|:-----:|:---------:|:----------:|:--------:|:---------:|:--------:|:---------:|:--------:|:---------:|:--------:|:---------:|\n| **Open-Source Models** | | | | | | | | | | | |\n| InternVL3.5-8B | | 0.40 | 35.6 | 0.41 | 1.8 | 0.51 | 69.4 | 0.38 | 52.9 | 0.30 | 35.0 |\n| InternVL3.5-A28B | | 0.56 | 58.1 | 0.54 | 28.5 | 0.69 | 85.5 | 0.56 | 63.3 | 0.46 | 75.2 |\n| Qwen2.5-VL-7B | | 0.44 | 34.8 | 0.54 | 8.0 | 0.62 | 17.0 | 0.42 | 36.4 | 0.21 | 90.5 |\n| Qwen2.5-VL-72B | | 0.49 | 57.2 | 0.59 | 18.0 | 0.66 | 91.5 | 0.46 | 56.6 | 0.26 | 86.0 |\n| Qwen3-VL-2B | | 0.57 | 35.2 | 0.61 | 5.5 | 0.71 | 11.8 | 0.50 | 37.9 | 0.42 | 93.0 |\n| Qwen3-VL-8B | | 0.66 | 60.9 | 0.69 | 32.5 | 0.77 | **97.2** | 0.64 | 59.1 | 0.56 | 81.0 |\n| Qwen3-VL-8B | ✓ | 0.49 | 45.9 | 0.52 | 11.2 | 0.64 | 79.7 | 0.51 | 53.4 | 0.32 | 56.2 |\n| Qwen3-VL-A22B | | 0.66 | 64.9 | 0.69 | 36.5 | 0.73 | 95.5 | 0.66 | 68.3 | 0.59 | 82.0 |\n| Qwen3-VL-A22B | ✓ | 0.65 | 60.4 | 0.67 | 31.0 | 0.75 | 93.5 | 0.65 | 62.3 | 0.54 | 78.0 |\n| Qwen3.5-A3B | | 0.71 | 68.1 | 0.79 | 36.8 | 0.81 | 84.2 | 0.68 | 75.6 | 0.57 | 84.2 |\n| Qwen3.5-A17B | | **0.73** | 72.2 | **0.81** | 52.0 | 0.81 | 81.3 | 0.67 | 75.3 | 0.66 | 89.4 |\n| Gemma 4 31B it | | 0.34 | 57.1 | 0.37 | 9.6 | 0.56 | 81.9 | 0.33 | 65.0 | 0.09 | 84.5 |\n| MiniCPM-V 4.5 | ✓ | 0.40 | 44.9 | 0.45 | 2.8 | 0.61 | 87.5 | 0.38 | 56.9 | 0.15 | 48.8 |\n| Molmo 7B-D 0924 | | 0.01 | 16.9 | 0.01 | **70.8** | 0.01 | 3.0 | 0.01 | 0.7 | 0.01 | 0.5 |\n| Molmo 72B 0924 | | 0.00 | 9.1 | 0.00 | 6.8 | 0.01 | 16.5 | 0.01 | 3.2 | 0.00 | 12.8 |\n| Ovis2.6-30B-A3B | ✓ | 0.53 | 39.7 | 0.54 | 8.5 | 0.63 | 77.9 | 0.57 | 71.6 | 0.42 | 12.2 |\n| GLM-4.5V 108B | ✓ | 0.44 | 56.6 | 0.45 | 11.5 | 0.61 | 84.5 | 0.44 | 63.3 | 0.23 | 81.5 |\n| Kimi K2.5 | | 0.71 | **77.0** | 0.73 | 70.2 | 0.78 | 78.2 | 0.72 | 77.8 | 0.66 | 86.0 |\n| Kimi K2.5 | ✓ | 0.70 | 72.3 | 0.75 | 68.5 | 0.78 | 81.7 | 0.60 | 65.3 | **0.66** | 84.8 |\n| **Proprietary Models** | | | | | | | | | | | |\n| GPT-4o | | 0.30 | 55.9 | 0.35 | 20.5 | 0.47 | 83.0 | 0.24 | 55.6 | 0.12 | 80.5 |\n| GPT-5 | | 0.38 | 62.1 | 0.50 | 36.2 | 0.57 | 59.6 | 0.21 | **78.1** | 0.18 | 71.0 |\n| Seed 1.8 | | 0.69 | 69.6 | 0.68 | 45.5 | 0.79 | 92.7 | 0.69 | 71.8 | 0.61 | 82.5 |\n| Seed 1.8 | ✓ | 0.67 | 71.1 | 0.69 | 48.0 | 0.78 | 89.2 | 0.57 | 73.3 | 0.60 | 80.8 |\n| Seed 2.0 Pro | | 0.72 | 76.1 | 0.75 | 60.8 | 0.81 | 82.0 | **0.73** | 77.6 | 0.62 | 92.2 |\n| Seed 2.0 Pro | ✓ | 0.71 | 75.3 | 0.76 | 61.8 | 0.80 | 82.0 | 0.65 | 74.3 | 0.66 | 89.0 |\n| MiMo-V2-Omni | ✓ | 0.56 | 62.3 | 0.62 | 40.0 | 0.71 | 80.7 | 0.58 | 73.3 | 0.36 | 64.2 |\n| Gemini 2.5 Pro | ✓ | 0.53 | 56.3 | 0.67 | 33.2 | 0.72 | 39.6 | 0.49 | 59.4 | 0.23 | 95.0 |\n| Gemini 3.1 Pro | ✓ | 0.70 | 73.1 | 0.80 | 61.0 | **0.83** | 62.7 | 0.66 | 71.1 | 0.52 | **95.8** |\n| Claude Opus 4.7 | ✓ | 0.50 | 66.8 | 0.53 | 50.2 | 0.63 | 74.4 | 0.44 | 56.6 | 0.38 | 86.0 |\n\n> **Cl** = Clerical, **Re** = Regular, **Ru** = Running, **Cu** = Cursive. **Bold** = best.\n\n## Getting Started\n\n### 1. Setup\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FVirtualLUOUCAS\u002FChronicles-OCR.git\ncd Chronicles-OCR\npip install -r requirements.txt\n```\n\n### 2. Download Data\n\nDownload and place the benchmark data under `data\u002F`:\n\n```\ndata\u002F\n├── Chronicles_OCR.jsonl\n└── images\u002F\n    ├── 甲骨文\u002F    # Oracle Bone\n    ├── 金文\u002F      # Bronze Script\n    ├── 篆书\u002F      # Seal Script\n    ├── 隶书\u002F      # Clerical Script\n    ├── 楷书\u002F      # Regular Script\n    ├── 行书\u002F      # Running Script\n    └── 草书\u002F      # Cursive Script\n```\n\n### 3. Inference\n\n```bash\n# OpenAI-compatible API\npython infer.py --api_type openai_compat \\\n    --model_name Qwen2.5-VL-7B-Instruct \\\n    --base_url http:\u002F\u002F127.0.0.1:8000\u002Fv1 \\\n    --api_key EMPTY --max_workers 64\n\n# Local vLLM\npython infer.py --api_type local_vllm \\\n    --model_path \u002Fpath\u002Fto\u002Fmodel \\\n    --tensor_parallel_size 1 --max_model_len 32768\n```\n\n### 4. Judging (Rule-based)\n\n```bash\npython judge.py                    # all models\npython judge.py --models model_a   # specific model\n```\n\n### 5. Summary Report\n\n```bash\npython summarize.py\n# → judge_results\u002Fresults_analysis.xlsx\n```\n\n## Citation\n\n```bibtex\n@misc{li2026chronicles,\n      title={Chronicles-OCR: A Cross-Temporal Perception Benchmark for the Evolutionary Trajectory of Chinese Characters},\n      author={Gengluo Li and Shangping Peng and Xingyu Wan and Chengquan Zhang and Hao Feng and Xin Xu and Pian Wu and Bang Li and Zengmao Ding and Yongge Liu and Yipei Ye and Yang Yang and Zhan Shu and Guojun Yan and Zhe Li and Can Ma and Weiping Wang and Yu Zhou and Han Hu},\n      year={2026},\n      journal={arXiv preprint arXiv:2605.11960},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.11960},\n}\n```\n\n## Acknowledgements\n\nWe sincerely acknowledge the Key Laboratory of Oracle Bone Inscription Information Processing at Anyang Normal University and the Palace Museum for their invaluable contributions to data sourcing and expert annotation.\n\n## License\n\nThis benchmark is released for **research purposes only**.\n","Chronicles-OCR 是一个专门用于评估视觉语言模型跨时间感知能力的基准，涵盖了汉字从甲骨文到楷书等七种主要书写形式的演化轨迹。该项目包含2800张严格平衡的图像，这些图像来源于多种物理媒介，如龟甲、青铜器和纸本书法。核心功能包括对不同历史时期汉字的识别与分析，旨在帮助研究人员理解并改进模型在处理古老至现代汉字时的表现。Chronicles-OCR 适合于需要深入研究汉字演变历程及其对现代文本识别技术影响的研究场景，同时也为开发更加鲁棒的文字识别系统提供了宝贵的数据支持。","2026-06-11 03:55:36","CREATED_QUERY"]