zsyjsld/neijing-sft-v3

Neijing SFT V3 Dataset (杏核)

This dataset is a specialized SFT fine-tuning dataset for the formalization, computational derivation, and clinical reasoning of Huangdi Neijing. It integrates classical Traditional Chinese Medicine (TCM) theories with system dynamics equations under a strict double-purity standard.

Dataset Details

• Total Entries: 570 pairs (Standard SFT)
• Data Categories:
  • Type B (370 Academic Scenarios): Covers 41 chapters of the Huangdi Neijing, with rigorous logical derivations.
  • Type A (50 Core Definitions): Defines Yin-Yang, Sheng Qi, and observation mappings inside the meta-framework.
  • Type C (150 Identity Definitions): Anchors the model identity self-cognition as the Xinghe (杏核) inference engine.

Double-Purity Standard

1. Thinking Chain (inside <think>): Implements strict parameter binding and mathematical derivation using the unified meta-framework:

$$\dot{x}=F(x)+G(x,\mu (t))\backslash dot\{x\}\; =\; F(x)\; +\; G(x,\; \backslash mu(t))$$

$$y=h(x)+ϵy\; =\; h(x)\; +\; \backslash epsilon$$

2. Clinical Output (outside <think>): Completely filters out 25 modern system science terms (such as "system, dynamics, control, equation") and English characters, producing natural clinical TCM explanations.

License

This dataset is released under the Apache 2.0 License.