新的 RAG 研究增强了 LLM 的检索、遗忘和忠实性

Apple Machine Learning Research TIER_1 English(EN) · 2026-07-15 00:00

CLaRa：利用连续潜在推理连接检索与生成

Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge but still suffers from long contexts and disjoint retrieval–generation optimization. In this work, we propose CLaRa (Continuous Latent Reasoning), a unified framework that performs …

arXiv cs.CL TIER_1 English(EN) · Shang Wang, Tianqing Zhu, Dayong Ye, Wanlei Zhou · 2026-07-17 04:00

当机器遗忘遇上检索增强生成（RAG）：保守秘密还是遗忘知识？

arXiv:2410.15267v3 Announce Type: replace-cross Abstract: The deployment of large language models (LLMs) like ChatGPT and Gemini has shown their powerful natural language generation capabilities. However, these models can inadvertently learn and retain sensitive information and h…

arXiv cs.CL TIER_1 English(EN) · Xin Sun, Zhongqi Chen, Qiang Liu, Shu Wu, Bowen Song, Weiqiang Wang, Zilei Wang, Liang Wang · 2026-07-15 04:00

预测检索！检索增强生成的时间自适应测试

arXiv:2601.11443v3 Announce Type: replace Abstract: Retrieval-Augmented Generation (RAG) has emerged as a powerful approach for enhancing large language models' question-answering capabilities through the integration of external knowledge. However, when adapting RAG systems to sp…

arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Tao Jia · 2026-07-13 06:51

NGM-RAG：基于神经图匹配的检索增强生成

Retrieval-Augmented Generation (RAG) significantly enhances the ability of Large Language Models (LLMs) to provide accurate and contextually relevant answers by dynamically integrating external databases. However, traditional RAG methods are primarily constrained by their relianc…

arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Zhihui Sun · 2026-07-11 13:43

SVD-RAG：通过奇异值分解实现高效的树状检索增强生成

Retrieval-Augmented Generation (RAG) systems enhance large language models by retrieving relevant documents from external knowledge bases. Recent work by Sarthi et al. (2024) introduced RAPTOR, which organizes documents into hierarchical tree structures for efficient retrieval, b…

arXiv cs.AI TIER_1 English(EN) · Pengcheng Huang, Zhenghao Liu, Yukun Yan, Haiyan Zhao, Xiaoyuan Yi, Hao Chen, Zhiyuan Liu, Maosong Sun, Tong Xiao, Ge Yu, Chenyan Xiong · 2026-07-10 04:00

ParamMute：抑制知识关键FFN以实现忠实的检索增强生成

arXiv:2502.15543v4 Announce Type: replace-cross Abstract: Large language models (LLMs) integrated with retrieval-augmented generation (RAG) have improved factuality by grounding outputs in external evidence. However, they remain susceptible to unfaithful generation, where outputs…

arXiv cs.IR (Information Retrieval) TIER_1 Deutsch(DE) · Qinyong Wang · 2026-07-09 11:22

DaV-Gen：通过草稿与验证实现端到端生成式检索

Mainstream industrial information retrieval systems (e.g., search and recommendation) are usually built upon Multi-Stage Cascade Architectures (MCAs), which balance effectiveness and efficiency through a coarse-to-fine ``retrieval-ranking'' pipeline. However, the optimization obj…

arXiv cs.CL TIER_1 English(EN) · Korbinian Randl, Guido Rocchietti, Aron Henriksson, Ziawasch Abedjan, Tony Lindgren, John Pavlopoulos · 2026-07-08 04:00

量化RAG中检索器-生成器对齐与局部解释

arXiv:2601.21803v2 Announce Type: replace Abstract: Retrieval-Augmented Generation (RAG) systems combine dense retrievers and language models to ground their outputs in external documents. However, the interaction between these components remains opaque, creating challenges for d…

arXiv cs.CL TIER_1 English(EN) · Yingdong Yang, Haijian Wu · 2026-07-07 04:00

面向LongEval-RAG的候选约束检索增强生成：系统设计与实证分析

arXiv:2607.04008v1 Announce Type: new Abstract: We present a candidate-constrained retrieval-augmented generation system for LongEval-RAG, where each query is associated with an organizer-provided candidate set and all retrieved evidence and final citations must remain within tha…

arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Haijian Wu · 2026-07-04 20:07

面向LongEval-RAG的长文本约束检索增强生成：系统设计与实证分析

We present a candidate-constrained retrieval-augmented generation system for LongEval-RAG, where each query is associated with an organizer-provided candidate set and all retrieved evidence and final citations must remain within that set. The system combines deterministic provena…

arXiv cs.AI TIER_1 English(EN) · Valentin J. J. Kreileder, Johannes Reisinger, Andreas Fischer · 2026-07-03 04:00

评估检索增强生成在学术文本上的分块策略

arXiv:2607.01852v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) systems use the question-answering capabilities of Large Language Models (LLMs) to access information outside their parameters. We evaluate if cluster-based semantic chunking improves retrieval…

arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Kirill Khrylchenko · 2026-07-02 23:13

面向大规模生成检索的离策略REINFORCE长期优化

Generative retrieval has become a popular paradigm for large-scale recommendation. However, it is typically trained with supervised next-item prediction objectives that do not directly optimize long-term user satisfaction. In this work, we formulate recommendation as a session-le…

Hugging Face Daily Papers TIER_1 English(EN) · 2026-07-02 08:12

评估检索增强生成在学术文本上的分块策略

Retrieval-Augmented Generation (RAG) systems use the question-answering capabilities of Large Language Models (LLMs) to access information outside their parameters. We evaluate if cluster-based semantic chunking improves retrieval and answer quality compared to fixed-size and rec…

arXiv cs.CL TIER_1 English(EN) · Andreas Fischer · 2026-07-02 08:12

评估检索增强生成在学术文本上的分块策略

Retrieval-Augmented Generation (RAG) systems use the question-answering capabilities of Large Language Models (LLMs) to access information outside their parameters. We evaluate if cluster-based semantic chunking improves retrieval and answer quality compared to fixed-size and rec…

arXiv cs.CL TIER_1 English(EN) · Raymond Li, Md Tawkat Islam Khondaker, Amirhossein Abaskohi, Gabriel Murray, Giuseppe Carenini, Issam H. Laradji · 2026-07-02 04:00

用于检索增强生成的双置信度对比解码

arXiv:2607.00570v1 Announce Type: new Abstract: Retrieval-augmented generation (RAG) increasingly requires models to answer questions from multiple retrieved documents, where only some sources are relevant and the retrieved bundle may contain stale, noisy, or conflicting evidence…

arXiv cs.CL TIER_1 English(EN) · Yushi Sun, Bowen Cao, Wai Lam · 2026-07-02 04:00

当经典缓存策略失效时：用于语义检索缓冲区的学习增强型替换

arXiv:2607.00394v1 Announce Type: cross Abstract: LLM agents increasingly rely on retrieval buffers to store and reuse past experience, yet the cache management policies governing these buffers remain largely ad-hoc. We formalize this as an online semantic cache replacement probl…

arXiv cs.AI TIER_1 English(EN) · Xue Tan, Yi Zheng, Chang Huo, Yunruo Zhang, Yu Liu, Hao Luan, Zhuyang Yu, Xiaoyan Sun, Ping Chen, Jun Dai · 2026-07-02 04:00

PRA-RAG：可证明的鲁棒聚合用于检索增强生成以对抗检索损坏

arXiv:2607.00012v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by incorporating external knowledge, effectively mitigating their inherent knowledge limitations. However, RAG remains vulnerable to poisoning attacks that…

arXiv cs.CL TIER_1 English(EN) · Issam H. Laradji · 2026-07-01 07:55

用于检索增强生成的双置信度对比解码

Retrieval-augmented generation (RAG) increasingly requires models to answer questions from multiple retrieved documents, where only some sources are relevant and the retrieved bundle may contain stale, noisy, or conflicting evidence. Existing contrastive decoding methods primaril…

arXiv cs.CL TIER_1 English(EN) · Wai Lam · 2026-07-01 03:38

当经典缓存策略失效时：用于语义检索缓冲区的学习增强型替换

LLM agents increasingly rely on retrieval buffers to store and reuse past experience, yet the cache management policies governing these buffers remain largely ad-hoc. We formalize this as an online semantic cache replacement problem with switching costs, where items are matched b…

arXiv cs.AI TIER_1 English(EN) · Ernesto Lopez Fune (DE) · 2026-06-30 04:00

高维浓度与嵌入空间中的检索不稳定性：对检索增强生成的影响

arXiv:2606.28330v1 Announce Type: cross Abstract: Embedding-based retrieval systems rely on the assumption that geometric proximity in highdimensional representation spaces reflects semantic relevance. However, high-dimensional geometry induces concentration phenomena that can re…

arXiv cs.CL TIER_1 English(EN) · Zhe Dong (University of Maine at Presque Isle), Fang Qin (Stanford University), Manish Shah (Independent Researcher), Yicheng Wang (Independent Researcher) · 2026-06-30 04:00

取用前须知：检索增强生成（RAG）的校准检索预算分配

arXiv:2606.29959v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) typically retrieves a fixed number of passages for every query. This is wasteful when the reader already knows the answer, and it can be harmful when irrelevant or partially relevant passages d…

arXiv cs.AI TIER_1 English(EN) · Ian van Dort (University of Amsterdam), Maria Heuss (University of Amsterdam) · 2026-06-30 04:00

大型语言模型如何引用？检索增强生成中归因的机制化解读

arXiv:2606.28358v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) aims to enhance the trustworthiness of Large Language Models (LLMs) by grounding their outputs in external documents, often using inline citations for verifiability. However, the faithfulness o…

arXiv cs.AI TIER_1 English(EN) · Qili Zhang, Qianren Mao, Yangyifei Luo, Yashuo Luo, Hanwen Hao, Zhilong Cao, Weifeng Jiang, Zhijun Chen, Junnan Liu, Feng Yan, Xiaolong Wang, Jinlong Zhang, Zhenting Huang, Zhixing Tan, Jie Sun, Bo Li, Jianxin Li, Philip S. Yu · 2026-06-30 04:00

XRAG: 探究核心 -- 评估高级检索增强生成中的基础组件

arXiv:2412.15529v4 Announce Type: replace-cross Abstract: Retrieval-augmented generation (RAG) synergizes the retrieval of pertinent data with the generative capabilities of Large Language Models (LLMs), ensuring that the generated output is not only contextually relevant but als…

arXiv cs.AI TIER_1 English(EN) · Bingxue Zhang, Jianying Jia, Feida Zhu · 2026-06-30 04:00

覆盖未知：检索增强生成的信息需求覆盖优化

arXiv:2606.29328v1 Announce Type: cross Abstract: Retrieval-augmented generation (RAG) typically treats context selection as ranking chunks against a single query embedding. This assumption breaks down for complex queries, such as multi-hop or ambiguous questions, where top-k sel…

arXiv cs.AI TIER_1 English(EN) · Ansh Kamthan · 2026-06-30 04:00

AB-RAG：用于可靠问答的自适应预算检索增强生成

arXiv:2606.29090v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) has become the standard way to ground large language models in external knowledge, yet most systems retrieve a fixed number of passages for every question regardless of its difficulty. This was…

arXiv cs.AI TIER_1 English(EN) · Bharath Simha Reddy Muthyam · 2026-06-30 04:00

检索增强生成中的敏感性、鲁棒性和稳定性的系统级分析

arXiv:2606.28337v1 Announce Type: cross Abstract: Retrieval-Augmented Generation (RAG) systems are often evaluated using final answer accuracy, even though their failures can originate from preprocessing, retrieval, context packing, or generation. This paper presents a controlled…

arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Yicheng Wang · 2026-06-29 08:36

取用前须知：检索增强生成（RAG）的校准检索预算分配

Retrieval-augmented generation (RAG) typically retrieves a fixed number of passages for every query. This is wasteful when the reader already knows the answer, and it can be harmful when irrelevant or partially relevant passages distract the reader. We formulate adaptive RAG as c…

arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Feida Zhu · 2026-06-28 10:49

覆盖未知：检索增强生成的信息需求覆盖优化

Retrieval-augmented generation (RAG) typically treats context selection as ranking chunks against a single query embedding. This assumption breaks down for complex queries, such as multi-hop or ambiguous questions, where top-k selection tends to over-cover one semantic aspect whi…

Hugging Face Daily Papers TIER_1 English(EN) · 2026-06-27 21:08

AB-RAG：用于可靠问答的自适应预算检索增强生成

Retrieval-Augmented Generation (RAG) has become the standard way to ground large language models in external knowledge, yet most systems retrieve a fixed number of passages for every question regardless of its difficulty. This wastes computation on easy questions, starves hard on…

arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Ansh Kamthan · 2026-06-27 21:08

AB-RAG：用于可靠问答的自适应预算检索增强生成

Retrieval-Augmented Generation (RAG) has become the standard way to ground large language models in external knowledge, yet most systems retrieve a fixed number of passages for every question regardless of its difficulty. This wastes computation on easy questions, starves hard on…

dev.to — MCP tag TIER_1 English(EN) · Kamalesh AR · 2026-07-21 15:41

我探索RAG（检索增强生成）的演示之旅

<p>Recently, I had the opportunity to deliver a presentation on Retrieval-Augmented Generation (RAG) at my institute. Preparing for this session helped me understand not only what RAG is, but also why it has become one of the most important techniques in modern AI applications.</…

Towards AI TIER_1 English(EN) · Srini Dwarakanathan · 2026-07-19 23:01

OpenSearch 生产 RAG 优化，第二部分：词汇检索

<p><em>This is Part 2 of a series on optimizing OpenSearch for production RAG. </em><a href="https://medium.com/towards-artificial-intelligence/opensearch-optimizations-for-production-rag-c335ac49f3e5"><em>Part 1 covered semantic retrieval</em></a><em>, meaning vector search with…

dev.to — MCP tag TIER_1 English(EN) · Giulio D'Erme · 2026-07-18 12:07

检索增强的自我回忆 — 第 6 部分：毫无作用的微调，以及将其作为 MCP 服务器发布

<p><em>Part 6 (finale) of Retrieval-Augmented Self-Recall. Code: <a href="https://github.com/GiulioDER/RE-call" rel="noopener noreferrer">RE-call</a>. <a href="https://dev.to/gde03/retrieval-augmented-self-recall-part-5-the-gap-threshold-that-didnt-transfer-86a">Part 5: the gap t…

Medium — fine-tuning tag TIER_1 English(EN) · Cibidarwin · 2026-07-07 06:44

什么是检索增强生成（RAG），它与微调有何不同？

<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@cibidarwin1996/what-is-retrieval-augmented-generation-rag-and-how-is-it-different-from-fine-tuning-853604d3873a?source=rss------fine_tuning-5"><img src="https://cdn-images-1.medium.com/max/260…

dev.to — MCP tag TIER_1 English(EN) · Keerat Rashid · 2026-07-06 15:47

使用 LangChain 和 Pinecone 构建检索增强生成 (RAG) 系统

<p>While LLMs are great, there are some limitations in using LLMs: LLMs can hallucinate, presenting factually incorrect information when they don't know the answers, and their knowledge gets frozen at the time of training. That's when Retrieval Augmented Generation (RAG) addresse…

Towards AI TIER_1 English(EN) · Anthony Usoro · 2026-07-06 05:25

什么是检索增强生成（RAG）？企业完整指南

<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*O4lEhDU-U9byg1vIhZpbFw.png" /><figcaption>RAG Image</figcaption></figure><h4>How to get accurate, source-backed answers out of AI — instead of confident guesses.</h4><p>If you’ve spent any time with ChatGPT, Clau…

dev.to — LLM tag TIER_1 Deutsch(DE) · Machine coding Master · 2026-07-23 05:39

告别朴素分块：Spring AI 中的晚期分块 RAG

<h2> Ditch Naive Chunking: Late Chunking RAG in Spring AI </h2> <p>Naive text chunking breaks your RAG pipeline by slicing context at arbitrary token boundaries before your embedding model ever sees the text. Late chunking fixes this structural flaw by running a long-context tran…

dev.to — LLM tag TIER_1 English(EN) · Imus · 2026-07-21 13:34

报道来源 [196]