Published January 26, 2026 | Version 2.0
Dataset Open

Buddhist Classics AI Translation Series Vol.12: Kham-Amdo Multi-Sectarian Collections ( 康巴、安多诸宗法汇 Version 2.0)

  • 1. Independent Research Collective

Description

This is **Volume 12** of the comprehensive *Buddhist Classics AI Translation Series*, featuring **Kham-Amdo Multi-Sectarian Collections** from the Qing Dynasty and modern era, primarily from **Sichuan Garzê (Kham)** and **Qinghai-Gansu (Amdo)** regions.

---

## ⚠️ Version Notice

**This is Version 1.2 (temporary release). Version 2.0 upgrade planned for mid-2026.**

- **Current status**: Functional bilingual Chinese-English translations
- **Upcoming improvements**: Enhanced quality, expanded annotations, additional cross-references
- **Recommendation**: For scholarly citation, please await Version 2.0

---

## ⚠️ Copyright Restrictions

**This volume excludes one encrypted sub-collection due to uncertain modern copyrights:**

- **Excluded content**: Modern works with unclear copyright status (post-1949)
- **Storage**: Privately archived on Zenodo (encrypted, not publicly accessible)
- **Future release**: May be published if copyright clarification obtained
- **Note**: This encrypted sub-collection is NOT included in the public download

**Public release includes**: ~60+ authors, 400+ volumes, 10,000+ texts, ~80-90 million Chinese characters

---

## About This Volume

### **Historical Context**

This collection represents the **"Golden Age of Qing Dynasty Tibetan Buddhism"** in Kham and Amdo regions (modern **Sichuan Garzê, Qinghai, Gansu**), primarily spanning:

- **Early Qing (1644-1720)**: Text compilation, Degé Printing House establishment
- **Mid-Qing (1720-1850)**: Nyingma revival, Rimé Movement emergence
- **Late Qing (1850-1912)**: Peak of non-sectarian integration (Jamgön Kongtrul, Jamyang Khyentse)
- **Modern additions**: Some Republican-era and contemporary commentaries

**Geographic Focus**:
- **Kham (康区)**: Garzê Prefecture (Baiyu, Dergé, Sershul counties)
- **Amdo (安多)**: Qinghai-Gansu border regions (Choni, Labrang areas)

**Why Qing Era?**
> "The Qing Dynasty marked a transition from translating Indian classics (8th-14th centuries) to **systematically integrating** and **annotating** accumulated teachings. This volume captures that synthesis."

---

### **Major Collections Included**

#### **1. Nyingma Tradition (宁玛派)**
- **Jigme Lingpa** (吉美林巴, 1729-1798): *Longchen Nyingthig* (龙钦心髓) founder
- **Minling Terchen**: Mindrolling Monastery lineage
- **Karma Chagme** (噶玛乔美, ~60 volumes): *Great Perfection* and *Mahāmudrā* integration

#### **2. Kagyu Tradition (噶举派)**
- **Palpung Monastery** (八邦寺): Karma Kagyu second seat
- **Shangpa Kagyu**: Khyungpo Neljor lineage
- **Nedo Kagyu**: Unique Kham sub-lineage

#### **3. Sakya Tradition (萨迦派)**
- **Je Rendawa Zhönu Lodrö** (杰·热达瓦·雄努洛追, Yuan-Ming era): Complete works
- **Note**: Included here for convenience (not strictly Kham/Amdo or Qing)

#### **4. Rimé Movement (利美运动, Non-sectarian)**
- **Jamgön Kongtrul Lodrö Thaye** (蒋贡康楚, 1813-1899): *Five Treasuries* compiler
- **Jamyang Khyentse Wangpo** (蒋扬钦哲旺波, 1820-1892): Key Rimé figure

#### **5. Regional Traditions**
- **Choni (卓尼)**: Gelug practice texts from Gansu-Qinghai border
- **Palyul (白玉)**: Nyingma monastery collections
- **Dzogchen (竹钦)**, **Shechen (雪谦)**: Major Kham centers

---

### **Cultural and Academic Significance**

**1. Ethnographic Record**:
> "Rituals (*sgrub thabs*) in this volume are not merely religious texts but **vivid ethnographic snapshots** of Qing-era nomadic life: seasonal migrations, livestock blessings, community festivals, and death rites."

**2. Philosophical Synthesis**:
- **Dzogchen-Mahāmudrā integration**: Unifying Nyingma "Great Perfection" with Kagyu "Great Seal"
- **Prāṇa-nāḍī theory**: Detailed inner yogas (wind-channel-essence practices)
- **Terma systematization**: Codifying "treasure revelations" into coherent systems

**3. Linguistic Preservation**:
- **Pre-10th century Tibetan**: Archaic vocabulary in terma texts
- **Sanskrit mantras**: Seed-syllable (*bīja*) glossaries
- **Local dialects**: Kham-Amdo colloquialisms

**4. Cross-Cultural Exchanges**:
- **Tang-Tibet connections**: References to Vimalamitra's travels (8th century)
- **Ming Dynasty interactions**: Karma Chagme's knowledge of Chan Buddhism and Confucianism (Vol.35)
- **Qing imperial support**: Choni Printing House patronage

---

### **Notable Features**

#### **Karma Chagme's 60-Volume Collection**
- **Scale**: Largest single-author work in this volume
- **Content**: Daily practice manuals, death rituals (*phowa*), dream yoga, intermediate state guidance
- **Unique**: References to **Chinese Buddhism** (Chan, Pure Land) and **Confucian ethics**
- **Practical**: Designed for nomadic practitioners (portable rituals)

#### **Jigme Lingpa's Legacy**
- **Innovation**: Simplified Dzogchen for lay practitioners
- **Transmission**: Established Nyingthig lineage in Sichuan Garzê
- **Influence**: Foundation for modern Nyingma education

#### **Rimé Movement Texts**
- **Goal**: Break down sectarian barriers
- **Method**: Compile practices from Nyingma, Kagyu, Sakya, Jonang, Shangpa, Zhijé
- **Result**: Created "meta-tradition" preserving endangered teachings

---

## Technical Specifications

- **Total size**: ~80-90 million Chinese characters
- **Text count**: 10,000+ texts (400+ volumes, 60+ authors)
- **Languages**: Tibetan source, Chinese-English parallel translations
- **File formats**: Plain text (.txt), Markdown (.md), compressed archives (.7z)
- **Encoding**: UTF-8
- **Translation models**: Gemini 2.0 (primary)

**Known Issues**:
- ⚠️ **Name confusion**: AI may mix up similar names (e.g., Candrakīrti/Candragomin, Dharmakīrti/Dignāga)
- ⚠️ **Mantra errors**: Seed-syllables (*bīja*) may have incorrect romanization or missing glosses
- ⚠️ **Date inaccuracies**: Tibetan calendar conversions (Rabjung cycles) often wrong—users must verify
- ⚠️ **Geographic names**: Modern place names may be inaccurate—cross-check with historical maps

**Translation Prompts**:
1. **Seed-syllables**: Display (Tibetan, Devanāgarī, romanization, Chinese gloss) in continuous format
2. **Dates/places** (from Karma Chagme Vol.38 onward): Add (Common Era year, modern place name) in parentheses—**but AI accuracy is low**, verify independently

---

## User Advisory

### **What This Translation Is**:
- ✅ **First comprehensive Kham-Amdo corpus in modern languages**
- ✅ **Ethnographic treasure**: Daily rituals, nomadic life, cultural practices
- ✅ **Philosophical synthesis**: Qing Dynasty integration of earlier teachings

### **What This Translation Is Not**:
- ❌ **Polished academic edition**: AI errors, inconsistent formatting
- ❌ **Practice manual**: Requires qualified lama guidance + empowerments
- ❌ **Complete collection**: Some texts excluded (copyright issues, data gaps)

**For Academic Use**:
- Always verify against Tibetan sources
- Retranslate critical passages with Claude 3.7+ or similar
- Cross-check names, dates, mantras independently

**For Practice**:
- Consult qualified teachers
- Do not rely solely on translations for deity yoga or tantric rituals
- Understand cultural/historical context

### Explanation for Version 1.2 of the Kham and Amdo Various Sects Dharma Collection

In the g2.0 translation editions of volumes such as 1, 2, 3, 5, 6, 7, 8, 11, 12, etc., produced between July and November 2025, there were cases of minor omissions (approximately 1%) where entire paragraphs were missing in translation. To address this issue, a dedicated program was developed for electronic collation and supplementary translation.

Scattered revisions (0.n proportion) have been upgraded. Starting from November 22, 2025, the entire version group's cloud drive will no longer be replaced; updates will only be uploaded to the archive at https://huggingface.co/ospx1u. The latest versions may not appear in the long-term link table and can only be found at https://huggingface.co/ospx1u — please check there yourself. In general, frequent upgrades for volumes 1-12 are fully listed at https://huggingface.co/datasets/ospx1u/buddhist-classics-vol1-12/tree/main.

The structural format of this volume has also been adjusted, primarily by distinguishing directories for the Gemini 2.0 and Claude versions. However, the Claude versions are not fully collected; most of the completed ones from past upgrades have been incorporated into other volumes, and the remaining ones await future upgrades to version 2.0 for completion.

Starting from version 1.2, this volume newly adds content from the Dudjom New Terma series. In the public volumes, the complete works of Dudjom Lingpa and the collected writings of Sera Khandro are included. In the non-public volumes, the collected writings of the Second Dudjom Dharma King Jigdral Yeshe Dorje and those of Shaza Sangye Dorje are included, totaling 57 volumes (ka). The translator and Tibetologist Dr. Li Jianing from Oxford provided OCR electronic editions. The original image files for these electronic editions come from materials previously published (and now removed) by the Buddha Education Foundation — these are the clearest among the various published image editions available. Currently, only the Gemini 2.0 versions have been produced. Many colleagues participated in the collection and comparison of these materials this time, contributing significant effort — sincere thanks to all!

Buddhist Classics AI Translation Series  
December 15, 2025

 

Explanation for Version 2.0 of the Collected Dharma Teachings from Kham and Amdo Regions

This Version 2.0 primarily provides Claude model versions for all content. This volume represents the largest Tibetan-to-Chinese translation among Volumes 1-12. The Chinese translations of both public and non-public versions amount to approximately 100 million characters per single translation pass. The Claude model performs particularly well in maintaining parallelism in verse texts. Given that many sections of Volume 12 are still regularly included in daily liturgical practices, despite the substantial length, all content has been retranslated. The public volume uses the claude-sonnet-4-5-20250929 API model, while the non-public volume employs the traditional claude 3.7 sonnet model collected through window-based dialogue. Our assessment indicates that the 4.5 model does not surpass the 3.7 model in terms of literary quality for Tibetan-to-Chinese translation.

This volume concludes the 2016-2026 Tibetan-to-Chinese translation plan for the Buddhist Classics AI Translation Series. From the initial purchase of physical books, to the subsequent shift toward downloading digital texts, through some disorganized early translations (2016-2018), to the actual discovery of minimally readable content (2023), and finally to the actual text translations (2024-2026), the project has spanned nearly a decade. There are indeed many texts that should have been included but were not. If collection and Chinese translation continue in the future, it will likely be from 2027 onwards. The Translation Series will take a hiatus to learn how to manage and, where possible, upgrade these rather substantial datasets.

The Translation Series has completed the collection and translation of approximately 50GB of hundreds of thousands of documents, successfully converting Buddhist scriptures from the three major linguistic traditions—Chinese, Tibetan, and Pali—into Chinese and English data documents. All Tibetan-to-Chinese translations have been done in duplicate, and a considerable portion of Buddhist literature has been translated into Tibetan, Pali, Japanese, and Korean. The collection includes most of the renowned works produced throughout the history of Buddhist civilization, achieving humanity's first AI ancient text translation project comparable in scale to the Complete Library of the Four Treasuries (Siku Quanshu).

The completion of Version 2.0 of this volume coincides with humanity stepping into the era of AGI-level artificial intelligence. The Translation Series is also evaluating how to utilize new technological capabilities to revise old volumes and translate new ones in this evolving landscape. Facing the challenges of massive data and limited human resources, future efforts will inevitably need to elevate the re-editing of document collections to the level of engineering architecture. From micro-level translation techniques to macro-level civilizational transmission, everything must return to code-based implementation—these are all issues that require further contemplation.

Concurrent with the publication of this volume, Volume 16 Pali, Japanese, and Korean Translations of Kangyur and Tengyur Exoteric Sutras and Treatises has already been released. Later in 2026, Volume 17 Tibetan, Pali, Japanese, and Korean Translations of Early Indigenous Literature from the Chinese Buddhist Canon may also be published. However, since these are not new Chinese translations, this hiatus announcement is included here together with them.

Gratitude to all colleagues who have provided assistance.

 

Buddhist Classics AI Translation Series  
2026.01.26

 

 

Note: The supplementary volume of this volume, Buddhist Classics AI Translation Series Volume 12: Collected Dharma Teachings from Kham and Amdo Regions, Modern Section, Version 2.0, is a non-public encrypted file. Its preservation here merely indicates that these files have been completed and archived concurrently.

Notes (Jinyu Chinese)

---

## 中文说明

### ⚠️ 版本与版权提示

**1. 版本计划**:
- 当前:1.2版
- 升级:2026年年中发布2.0版

**2. 版权限制**:
- **不含**:现代版权不明文献(已加密存档在Zenodo,不公开)
- **包含**:清代及民国公有领域文献

### 三大特点

**1. 地理范围**:
- 康区(四川甘孜:白玉、德格、石渠)
- 安多(青海甘肃交界:卓尼、拉卜楞)

**2. 时代跨度**:
- 主要:清代(1644-1912)
- 少量:元明、民国

**3. 教派涵盖**:
- 宁玛(吉美林巴、噶玛乔美)
- 噶举(八邦寺、香巴、内多)
- 萨迦(杰·热达瓦)
- 利美(蒋贡康楚、蒋扬钦哲)

### 规模

- **作者**:60+位
- **卷帙**:400+函
- **篇目**:10000+篇
- **字数**:约8000-9000万汉字

### 核心价值

**1. 民族志记录**:
- 清代牧区生活(节庆、仪式、迁徙)
- 人畜祈福、超度亡者等实际操作

**2. 哲学综合**:
- 大圆满-大手印融合
- 气脉明点理论(那洛六法等)
- 伏藏系统化(敏珠林、噶玛乔美)

**3. 跨文化交流**:
- 噶玛乔美涉及禅宗、儒家(卷35)
- 唐蕃古道历史痕迹
- 明清皇室互动

### 特别推荐

**噶玛乔美60卷文集**:
- 本卷最大单一作者著作
- 日常修持手册(放牧间隙可修)
- 涉及汉传佛教和儒家

**吉美林巴**:
- 龙钦心髓创始人
- 简化大圆满(适合在家众)

**利美运动文献**:
- 打破壁垒
- 保存濒危教法

### 使用警告

**AI错误高发区**:
- ⚠️ **人名混淆**:月称/月官、法称/陈那
- ⚠️ **咒语错误**:种子字拟音、梵文缺失
- ⚠️ **日期不准**:藏历换算常错
- ⚠️ **地名错误**:古今对应需核实

**解决办法**:
- 论文用途:必须核对藏文原文
- 重译关键段:用Claude 3.7+
- 自行验证:人名、日期、地名

康巴、安多诸宗法汇1.2版说明


1.2.3.5.6.7.8.11.12等卷2025年7月-11月制作的g2.0翻译本中有约1%弱数量,整段脱译
的情况,为这种情况做了专门的程序,进行了电子校勘和补译。
零散校勘(0.n比例)升级,自20251122起不再更替整个版本群的网盘,只上传到
https://huggingface.co/ospx1u 存档。  最新版本有可能不在长期链接表,只
在 https://huggingface.co/ospx1u 请自行查询,一般情况1-12卷的频繁升级全表
在 https://huggingface.co/datasets/ospx1u/buddhist-classics-vol1-12/tree/main

本卷的结构形态也进行了调整,主要是区分了gemini2.0和claude版本的目录,但是claude
版本的收集不全,主要是在过去升级中计入其他卷的已完成版本,余下的还有待于以后升级到
2.0版的时候补全。

本卷在1.2版起新增敦珠新伏藏系列的内容,在公开卷中列入了敦珠林巴全集和色拉康卓文集,
在不公开卷中列入了二世敦珠法王吉扎益西多吉文集和夏札桑吉多杰文集,共57函,译师、藏学家
牛津李佳宁博士提供了OCR电子版,这些电子版的原始图档来自佛陀教育基金会曾公布并已下架的
资料,是已公布的若干种图版中最清晰的。目前也只制作了gemini2.0的版本,本次围绕这些资
料的收集和比对有不少同仁参与,付出努力,在这里一并致谢!


                                                            佛典AI译丛
                                                            2025年12月15日

 

康巴、安多诸宗法汇2.0版说明

本卷2.0版主要提供了全部内容的claude模型版本。本卷是过去1-12卷里藏译中文献最大的一卷,
公开和不公开的版本的汉译,单一次汉译可能在一亿字左右,claude模型在颂文的对仗上做
的比较好,考虑到第12卷的很多部分是现在仍然在经常性排入功课的,所以这一版本虽然篇幅
大,仍然都做了重译。其中公开卷采用claude-sonnet-4-5-20250929 API 模型 , 不公开卷
采用了传统的视窗对话体所采集的claude3.7sonnet模型。使用的评估是,4.5模型在藏译汉
的文辞观感上并不优于3.7模型。本卷是佛典AI译丛2016-2026年度藏译汉规划的最后一卷,
译丛从一开始购买书籍,到转而下载文字版,有一些凌乱的早期翻译本(2016-2018),到实
际的发现有些起码可读的内容(2023),到实际的文本译出(2024-2026),也经历了差不多
十年。确实有很多应收未收的部分,如果未来继续收集并译汉,可能也要2027年以后,译丛将
休整一段时间,并学习怎么管理或在可能的情况下如何升级这些较为庞大的数据集。

译丛完成了约50G数十万计的文献的收集与翻译,成功将汉、藏、巴三大语系佛教典籍统一转化
为汉语和英语数据文献,其中藏译汉均做了双重翻译,并向藏、巴、日、韩翻译了相当部分的佛
教文献,收录了历史上佛教文明所产生的大部分名篇,实现了人类历史上首个相当于《四库全书》
规模的AI古籍翻译工程。这一卷的2.0版的译出,实际已经是人类一步踏入AGI类人工智能体的
时代,译丛也在评估在这种时代演进中如何使用新的技术条件来修旧卷和译新卷。面对海量数
据与人力资源的匮乏,未来必然需要将文献群的再编辑上升至工程架构层面。从微观翻译技巧到
宏观文明传承,都需要回归代码实现,这些都是有待思考的课题。

在本卷刊出的同时,已公布第16卷 巴利日韩译甘珠尔丹珠尔显教经论,在2026年晚些时候还可
能出第17卷 藏巴利日韩译汉文大藏经早期本土文献,但这些都不是新译成汉语的,所以这个休
整说明就一并列在这里。

感谢所有提供帮助的同仁。

                                                            佛典AI译丛
                                                            2026年01月26日

 
https://github.com/Buddhist-Classics-AI-Translation-Series/Buddhist-translations

 

注意: 本卷的副卷 佛典AI译丛第十二卷:康巴、安多诸宗法汇近代篇2.0版 是不公开的加密文件,保存在这里只是意味着这些文件已经做了,并同期存档。

Files

Files (480.4 MB)

Additional details

Dates

Collected
2026-01-26