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.
slm c2c peft continual-learning catastrophic-forgetting cku kv-cache llm mechanistic-interpretability small-language-models ai-architecture modular-llm llm-post-training multi-expert neurips-2026 unified-ai cache-to-cache deletion-tuning deleting-weights sparse-masks
-
Updated
Nov 13, 2026