A resource repository for machine unlearning in large language models
-
Updated
Aug 6, 2026
A resource repository for machine unlearning in large language models
Implementation for NeurIPS'24 paper "Black-Box Forgetting."
人脑式上下文压缩逻辑,大幅减少长任务上下文占用以及token消耗
Adaptive memory for AI. A DeepSeek Harness plugin prototype for selective recall, configurable context budgets, and history compaction, designed to reduce token usage while preserving conversational continuity.
适配调用JEPA模型的插件,提供一个本地的JEPA模型项目(需要进一步训练持续学习,已完成基本工具调用),可在家用计算机配置上运行。A plugin adapted to call the JEPA model, providing a local JEPA model project (requires further training for continual learning; basic tool calling has been completed) that can run on a home computer CPU.
To associate your repository with the selective-forgetting topic, visit your repo's landing page and select "manage topics."