诸神缄默不语的论文阅读笔记和分类

诸神缄默不语-个人技术博文与视频目录

legalAI+文本摘要算文本摘要，legalAI+其他算legalAI。

Re后面的顺序是我写笔记的顺序（这个是我之前写的顺序，现在我决定慢慢把这个序号给删掉了），论文本身的发表顺序标在论文标题前面了。

GNN

通用同质图节点表征

1. (2018 ICLR) Re37：读论文 G2G Graph2Gauss Deep Gaussian Embedding of Graphs: Unsupervised Inductive Learning via Rank
2. (2019 ICLR) Re0：读论文 PPNP/APPNP Predict then Propagate: Graph Neural Networks meet Personalized PageRank：端到端，先transform再propagate
3. (2019 KDD) Re3：读论文 PGE A Representation Learning Framework for Property Graphs
4. (2020 KDD) Re46：读论文 DAGNN Towards Deeper Graph Neural Networks
5. (2020 ICLR) Re2: 读论文 CS-GNN Measuring and Improving the Use of Graph Information in Graph Neural Networks
6. (2021 ICLR) Re1：读论文 C&S (Correct and Smooth) Combining Label Propagation and Simple Models Out-performs Graph Ne：解耦transform和propagate，再加一个correct

通用异质图节点表征

1. (2017 KDD) Re31：读论文 metapath2vec: Scalable Representation Learning for Heterogeneous Networks：异质图版的node2vec
2. (2020 AAAI 滴滴+北大) Re22：读论文 HetSANN An Attention-based Graph Neural Network for Heterogeneous Structural Learning
3. (2021 KDD 清华) Re10：读论文 Are we really making much progress? Revisiting, benchmarking, and refining heterogeneous gr：喷了一圈各种HGNN算法，最后提出了一个简单的HGNN模型然后发现这个新模型表现最好了

同质图链路预测

1. (2020 IJCAI) 论文阅读笔记：DEAL_inductive链路预测_分别表征节点特征和拓扑结构+对比学习对齐：拓扑表征和特征表征分开建模
   论文全名：Inductive Link Prediction for Nodes Having Only Attribute Information

基础模型

RNN

1. (2014 NIPS) Re71：读论文 Sequence to Sequence Learning with Neural Networks

LLM

1. (2018 ACL) Re73 读论文：ULMFiT Universal Language Model Fine-tuning for Text Classification
2. (2019 NAACL) Re63：读论文 BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
3. (2019 NeurIPS) Re72：读论文 XLM Cross-lingual Language Model Pretraining
4. (2020 EMNLP) Re55：读论文 Entities as Experts: Sparse Memory Access with Entity Supervision：将实体表征结合到LLM中
5. (2020 JMLR) Re70：读论文 T5 Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
6. (2022 谷歌) Re69：读论文 LaMDA: Language Models for Dialog Applications
7. (2022 ICLR 谷歌) Re68：读论文 instruction tuning FLAN Finetuned Language Models Are Zero-Shot Learners
8. (2023 Meta) Re75 读论文：Toolformer: Language Models Can Teach Themselves to Use Tools

GPT系

1. (2018 OpenAI) GPT-1论文阅读笔记_Improving Language Understanding by Generative Pre-Training
2. (2019 OpenAI) GPT-2论文阅读笔记_Language Models are Unsupervised Multitask Learners
3. (2020 NeurIPS OpenAI) Re65：读论文 GPT-3 Language Models are Few-Shot Learners
4. (2023) Re78 读论文：GPT-4 Technical Report

Claude系

1. 论文阅读笔记：Claude如何思考
   (2025) On the Biology of a Large Language Model
   (2025) Circuit Tracing: Revealing Computational Graphs in Language Models

DeepSeek系

1. (2025) Re 80 读论文：DeepSeek-V3：2025年初最强大模型（几天前还是的）
2. (2025) Re 83 读论文：DeepSeek-R1：2025年初最强大模型的推理时候

Qwen系

1. (2025) Re 82：读论文：qwen 3

继续预训练

1. (2020 ACL) Re26：读论文 Don’t Stop Pretraining: Adapt Language Models to Domains and Tasks：继续预训练能提升下游任务表现效果

部分微调

1. Re77 读论文：LoRA: Low-Rank Adaptation of Large Language Models

prompt

1. (2022 ACM Computing Surveys 卡耐基梅隆大学) Re33：读论文 Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Languag：prompt综述
2. (2023 EMNLP) Re 84 读论文：TELeR: A General Taxonomy of LLM Prompts for Benchmarking Complex Tasks：prompt分类

位置编码

1. (2021 苏剑林) Re 79 读论文：RoPE RoFormer: Enhanced Transformer with Rotary Position Embedding

LM中蕴含的知识

1. (2019 EMNLP) Re51：读论文 Language Models as Knowledge Bases?：完形填空
2. (2020 EMNLP) Re52：读论文 How Much Knowledge Can You Pack Into the Parameters of a Language Model?：QA
3. (2020 TACL) Re53：读论文 How Can We Know What Language Models Know?：完形填空，但是template是通过学习得到的
4. (2023 EMNLP) Re67：读论文 Don‘t Trust ChatGPT when Your Question is not in English: A Study of Multilingual Abilities

RAG

叠实体表征那几个真的能算检索增强吗？
不过既然别人这么分类我就姑且这么算了

1. (2020 ICLR 斯坦福+Facebook) Re48：读论文 kNN-LMs Generalization through Memorization: Nearest Neighbor Language Models：在语言模型计算出的token概率的基础上，增加kNN token概率
   kNN概率的计算：首先构建海量文本向量数据库，key是上下文表征，value是target token。对每个测试样本，用FAISS检索得到k个最近的样本，其target token的概率与向量距离成反比（向量越近，概率越大）
2. (2020 ICML 谷歌) Re58：读论文 REALM: Retrieval-Augmented Language Model Pre-Training：从海量维百中检索相关文本，然后加到输入文本后面。这个检索器是端到端预训练的
3. (2020 NeurIPS Facebook) Re59：读论文 Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
4. (2020 AKBC) Re54：读论文 How Context Affects Language Models‘ Factual Predictions：对比了不同检索方案。其实做得挺单薄的，也就是拿TF-IDF检索维基百科的检索器加上跟监督算法比较了一下。关于seperation的实验比较有参考价值
5. (2021 NAACL 谷歌) Re60：读论文 FILM Adaptable and Interpretable Neural Memory Over Symbolic Knowledge：这篇也是叠实体表征
6. (2022 ICLR 谷歌) Re57：读论文 Mention Memory: incorporating textual knowledge into Transformers through entity mention at：mention memory构建语料中的mention表征向量，TOME模型在实现下游任务时结合实体mention表征向量做sparse attention
7. (2023 ACL) Re49：读论文 When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parame：自适应选择少见实体加检索
8. (2023 ICML) Re50：读论文 Large Language Models Struggle to Learn Long-Tail Knowledge：检索增强解决LM搞不动长尾实体的问题
9. (2024 谷歌) Re74 读论文：DataGemma Knowing When to Ask - Bridging Large Language Models and Data

agent

1. Re 81 读论文：AlphaEvolve: A Gemini-powered coding agent for designing advanced algorithms

AI编程

1. 论文阅读笔记：AI编程用的AGENTS.md应该不写或少写
   论文：(2026) Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?

文本摘要

1. (2017 ACL) Re47：读论文PGN/Pointer-Generator Netwoks Get To The Point：Summarization with Pointer-Generator Networks：生成概率+指针概率（抽取）
2. (2018 EMNLP) Re66：读论文 Bottom-Up Abstractive Summarization
3. (2019 EMNLP) Re64：读论文 BertSum Text Summarization with Pretrained Encoders & Fine-tune BERT for Extractive Summari
4. (2021 AAAI) Re4：读论文 CGSum: Enhancing Scientific Papers Summarization with Citation Graph：结合文献引用关系实现文献摘要
5. (2021 ACL 清华) Re5：读论文 TWAG: A Topic-guided Wikipedia Abstract Generator：结合维基百科的小标题生成摘要
6. (2022 AAAI) Re12：读论文 Se3 Semantic Self-segmentation for Abstractive Summarization of Long Legal Documents in Low：分治生成摘要
7. (2022 SIGIR) Re32：读论文 Summarizing Legal Regulatory Documents using Transformers
8. (2022 COLING 匹兹堡大学) Re35：读论文 ArgLegalSumm: Improving Abstractive Summarization of Legal Documents with Argument Mining：识别出argumen，然后生成摘要

LegalAI

1. (2020 ACL) Re23：读论文 How Does NLP Benefit Legal System: A Summary of Legal Artificial Intelligence：综述

LJP

1. (2017 EMNLP 北大) Re7：读论文 FLA/MLAC/FactLaw Learning to Predict Charges for Criminal Cases with Legal Basis：结合法条预测罪名
2. (2019 Law in Context) Re56：读论文 A Brief History of the Changing Roles of Case Prediction in AI and Law：美国LJP传统方法综述
3. (2020 ACL) Re27：读论文 LADAN Distinguish Confusing Law Articles for Legal Judgment Prediction：结合法条相似关系
4. (2021 ACL) Re16：读论文 ILDC for CJPE: Indian Legal Documents Corpus for Court Judgment Prediction and Explanation
5. (2021 NAACL 北大) Re18：读论文 GCI Everything Has a Cause: Leveraging Causal Inference in Legal Text Analysis
6. (2021 SIGIR 北大+阿里) Re21：读论文 MSJudge Legal Judgment Prediction with Multi-Stage Case Representation Learning in the Real
7. (2021 SIGIR) Re38：读论文 NeurJudge: A Circumstance-aware Neural Framework for Legal Judgment Prediction：结合犯罪情节
8. (2022 AAAI) 论文阅读笔记：LeSICiN_以inductive链路预测范式解决多标签文本分类任务（法条预测）：结合案例引用和法条层级异质图，用链路预测范式做法条预测任务
   论文全名：LeSICiN: A Heterogeneous Graph-based Approach for Automatic Legal Statute Identification from Indian Legal Documents
9. (2022 AAAI) Re14：读论文 ILLSI Interpretable Low-Resource Legal Decision Making
10. (2022 ACL 南大) Re11：读论文 EPM Legal Judgment Prediction via Event Extraction with Constraints：结合事件抽取
11. (2022 IJCAI 西电) Re28：读论文 CECP Charge Prediction by Constitutive Elements Matching of Crimes：结合犯罪要素+强化学习
12. (2022 IPM) Re36：读论文 CEEN Improving legal judgment prediction through reinforced criminal element extraction：结合犯罪要素+强化学习
13. (2022 COLING) Re 39：读论文 CTM Augmenting Legal Judgment Prediction with Contrastive Case Relations：结合案例标签相似关系和频率
14. (2022 Artificial Intelligence and Law) Re41：NumLJP Judicial knowledge‑enhanced magnitude‑aware reasoning for numerical legal judgment predi：结合数值信息

分类

1. (2022 NAACL) Re29：读论文 D2GCLF: Document-to-Graph Classifier for Legal Document Classification

文本相似性

1. (2020 SIGIR) Re8：读论文 Hier-SPCNet: A Legal Statute Hierarchy-based Heterogeneous Network for Computing Legal Case：结合案例引用和法条层级异质图

案例匹配

1. (2022 SIGIR 人大+华为) Re24：读论文 IOT-Match Explainable Legal Case Matching via Inverse Optimal Transport-based Rationale Ext

案例检索/推荐

1. (2022 SIGIR) Re25：读论文 Lecut+JOTR Incorporating Retrieval Information into the Truncation of Ranking Lists in the
2. (2022 ACM Transactions on Information Systems 清华+IBM) Re30：读论文 LegalGNN: Legal Information Enhanced Graph Neural Network for Recommendation

公平性

1. (2022 AAAI) Re13：读论文 Gender and Racial Stereotype Detection in Legal Opinion Word Embeddings

事件检测

1. (2022 ACL 清华) Re15：读论文 LEVEN: A Large-Scale Chinese Legal Event Detection Dataset

信息抽取

1. (2021 ACL) Re17：读论文 Challenges for Information Extraction from Dialogue in Criminal Law
2. (2021 NAACL) Re19：读论文 Paragraph-level Rationale Extraction through Regularization: A case study on European Court

其他

1. (2021 NAACL 剑桥) Re20：读论文 What About the Precedent: An Information-Theoretic Analysis of Common Law
2. (2022 SIGIR) Re34：读论文 Organizing Portuguese Legal Documents through Topic Discovery

信息抽取

意图识别+槽填充

1. (2021 ACL) Re 40：读论文 GL-GIN: Fast and Accurate Non-Autoregressive Model for Joint Multiple Intent Detection and

数值推理

MWP

1. (2014 EMNLP) Re42：读论文 ARIS Learning to Solve Arithmetic Word Problems with Verb Categorization：识别和分类动词
2. (2017 EMNLP) Re43：读论文 DNS Deep Neural Solver for Math Word Problems：RNN+检索
3. (2021 OpenAI) Re44：数据集 GSM8K 和 论文 Training Verifiers to Solve Math Word Problems：verifier
4. (2024 AAAI) Re61：读论文 PRP Get an A in Math: Progressive Rectification Prompting

向量压缩

向量量化

1. 论文阅读笔记：TurboQuant_谷歌整的新活正在攻击你的内存板块美股，还有造假和抄袭嫌疑
   (2026 ICLR 谷歌) TurboQuant: Online Vector Quantization with Near-optimal Distortion Rate
   涉及争议：(2024 SIGMOD) RaBitQ: Quantizing High-Dimensional Vectors with a Theoretical Error Bound for Approximate Nearest Neighbor Search

交叉学科

情报学

1. (2022 图书情报工作) Re76 读论文：新兴技术的多指标量化识别研究——基于向量表征方法的探索