大模型持续学习论文库
本目录包含原始 PDF、由 pdf-paper-reader 工作流生成的 Obsidian/PDF++ 精读笔记，以及可复核的文本与图表坐标。
当前共收录 41 篇。第 35–41 篇聚焦长期自适应智能所需的快速状态、测试时学习、神经记忆、记忆治理、离线巩固与元可塑性。
| # | 论文 | 发表 | 笔记 | 原文 |
|---|---|---|---|---|
| 01 | Continual Learning of Large Language Models: A Comprehensive Survey | 2024 · arXiv | 精读笔记 | |
| 02 | Continual Learning for Large Language Models: A Survey | 2024 · arXiv | 精读笔记 | |
| 03 | Continual Learning of Large Language Models | 2025 · EMNLP Tutorial | 精读笔记 | |
| 04 | Lifelong Learning of Large Language Model Based Agents: A Roadmap | 2025 · arXiv | 精读笔记 | |
| 05 | Continual Learning in Large Language Models: Methods, Challenges, and Opportunities | 2026 · arXiv | 精读笔记 | |
| 06 | Continual Learning and Catastrophic Forgetting | 2024 · arXiv | 精读笔记 | |
| 07 | Overcoming Catastrophic Forgetting in Neural Networks | 2017 · PNAS | 精读笔记 | |
| 08 | Gradient Episodic Memory for Continual Learning | 2017 · NeurIPS | 精读笔记 | |
| 09 | Lifelong Pretraining: Continually Adapting Language Models to Emerging Corpora | 2021 · arXiv | 精读笔记 | |
| 10 | ELLE: Efficient Lifelong Pre-training for Emerging Data | 2022 · Findings of ACL | 精读笔记 | |
| 11 | How to (Re)warm Your Model? Reusing Optimizer State for Continual Pre-Training | 2023 · arXiv | 精读笔记 | |
| 12 | Investigating Continual Pretraining in Large Language Models | 2024 · arXiv | 精读笔记 | |
| 13 | TiC-LM: A Multi-Year Benchmark for Continual Pretraining of Language Models | 2025 · ACL | 精读笔记 | |
| 14 | Continual Pre-training of Language Models: How to Re-warm Your Model? | 2025 · ACL | 精读笔记 | |
| 15 | Continual Learning via Sparse Memory Finetuning | 2025 · arXiv | 精读笔记 | |
| 16 | SCALE: Scalable Continual Learning via Model Expansion | 2026 · Findings of ACL | 精读笔记 | |
| 17 | TRACE: A Comprehensive Benchmark for Continual Learning in Large Language Models | 2023 · arXiv | 精读笔记 | |
| 18 | Orthogonal Subspace Learning for Language Model Continual Learning | 2023 · Findings of EMNLP | 精读笔记 | |
| 19 | ConPET: Continual Parameter-Efficient Tuning for Large Language Models | 2023 · arXiv | 精读笔记 | |
| 20 | InsCL: A Data-efficient Continual Learning Paradigm for Fine-tuning Large Language Models with Instructions | 2024 · NAACL | 精读笔记 | |
| 21 | SAPT: A Shared Attention Framework for Continual Learning of Large Language Models | 2024 · ACL | 精读笔记 | |
| 22 | Revisiting Catastrophic Forgetting in Large Language Model Tuning | 2024 · Findings of EMNLP | 精读笔记 | |
| 23 | Online Continual Learning of Large Language Models | 2025 · ICML | 精读笔记 | |
| 24 | ASO-LoRA: Adaptive Subspace Orthogonal LoRA for Continual Learning | 2026 · ACL | 精读笔记 | |
| 25 | WISE: Rethinking the Knowledge Memory for Lifelong Model Editing of Large Language Models | 2024 · NeurIPS | 精读笔记 | |
| 26 | MEMOIR: Lifelong Model Editing with Residual Memory | 2025 · NeurIPS | 精读笔记 | |
| 27 | LifelongAgentBench: Evaluating LLM Agents as Lifelong Learners | 2025 · arXiv | 精读笔记 | |
| 28 | Continual Learning Bench: Benchmarking Continual Learning for LLM Agents | 2026 · arXiv | 精读笔记 | |
| 29 | SkillLearnBench: Evaluating Continual Skill Learning in Language Agents | 2026 · arXiv | 精读笔记 | |
| 30 | LifeSkill: Lifelong Skill Learning for Language Agents | 2026 · arXiv | 精读笔记 | |
| 31 | Nested Learning: The Illusion of Deep Learning Architecture | 2025 · NeurIPS | 精读笔记 | |
| 32 | Self-Adapting Language Models | 2025 · arXiv | 精读笔记 | |
| 33 | Learning, Fast and Slow: Towards LLMs That Adapt Continually | 2026 · arXiv | 精读笔记 | |
| 34 | Knowledge Circuits in Pretrained Transformers | 2025 · arXiv | 精读笔记 | |
| 35 | Using Fast Weights to Attend to the Recent Past | 2016 · arXiv | 精读笔记 | |
| 36 | Learning to (Learn at Test Time): RNNs with Expressive Hidden States | 2024 · arXiv | 精读笔记 | |
| 37 | Titans: Learning to Memorize at Test Time | 2025 · arXiv | 精读笔记 | |
| 38 | MemOS: A Memory OS for AI System | 2025 · arXiv | 精读笔记 | |
| 39 | Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories | 2026 · arXiv | 精读笔记 | |
| 40 | Organizing Memories for Generalization in Complementary Learning Systems | 2023 · Nature Neuroscience | 精读笔记 | |
| 41 | Meta-Learning Biologically Plausible Plasticity Rules with Random Feedback Pathways | 2023 · Nature Communications | 精读笔记 |
辅助文件
paper-manifest.json：题名、分类与公开来源。publication-dates.json：发表年份与 venue。reading-briefs.json：逐篇批判性阅读提纲。.items/*.json：与 PDF.js 对齐的文本项坐标映射。reading-audit.json：全库引用坐标与图表裁剪审计。
坐标说明
精读笔记中的 selection 与 rect 均指向同目录本地 PDF。坐标由本地文本层和渲染结果生成；升级 PDF 版本后应重新生成 .items 并运行校验。