This is the official repository for the project:
"Semantic Bundle AI: Stable Coordinate Semantics as a Complementary Layer to Large Language Models"
The full proposal is available on SSRN:
https://ssrn.com/abstract=6139446
We propose a novel "Complementary Layer" for LLMs that uses a stable coordinate system to mitigate "Embedding Drift" and reduce re-training costs.
The theoretical framework, mathematical formulations, and documentation provided in this project are licensed under:
Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0)
- Attribution: Give credit to Makoto Saitou.
- NonCommercial: No commercial use without a separate agreement.
- NoDerivatives: No modified versions allowed.
Commercial Inquiries: For implementation details, partnership, or commercial licensing, please contact me through GitHub Issues or via the contact info on my SSRN profile.
If you have questions, feedback, or interest in collaboration, please feel free to:
- Open an Issue: For public discussions and technical questions.
- **Direct Contact: m.saitou@ontheshouldersofgiants.jp