Modular language models via a fixed-dimension ABI bottleneck — domain modules are bit-exactly portable across model sizes
-
Updated
Nov 14, 2026 - Python
Modular language models via a fixed-dimension ABI bottleneck — domain modules are bit-exactly portable across model sizes
CKU Theory: Inspired by 'The Surprising Effectiveness of Deleting Weights in LLM Post-Training' (NeurIPS 2026) - Connecting task-specific Deletion Masks on one shared AI backbone via Cache-to-Cache (C2C) into One Unified AI, by AXNli Inc.
UXI-LLM is a modular hybrid language model that combines neural and symbolic reasoning for enhanced AI capabilities. Built for flexibility and local use, it empowers developers and researchers to customize their AI solutions with ease. 🐙💻
To associate your repository with the modular-llm topic, visit your repo's landing page and select "manage topics."