https://ym-xu.github.io/living-survey/fields/rag/#tag=efficiencyCompression, caching and the trade-off with long-context models.Thu, 12 Nov 2026 15:12:01 +0000https://arxiv.org/abs/2611.13675https://arxiv.org/abs/2611.13675Tue, 10 Nov 2026 00:00:00 +0000Senhao Liu, Yuheng Zhang, Chunyu Wei, Yueguo Chen, et al. · NeurIPS 2026 Retrieval-augmented generation (RAG) is bottlenecked by the token budget of large language models: feeding long documents into the context window is expensive, while aggressive textual compression discards fine-grained information that is irreversibly lost in a one-dimensional token sequence. We argue that this bottlen…https://arxiv.org/abs/2611.10657https://arxiv.org/abs/2611.10657Sat, 07 Nov 2026 00:00:00 +0000Haoran Luo, Ziyue Zhu, Shangyang Wu, Linhao Luo, et al. · NeurIPS 2026 Driven by increasingly complex real-world applications, Retrieval-Augmented Generation (RAG) has evolved from textual pipelines to multimodal settings, comprising two complementary dimensions: knowledge organization and retrieval enhancement. However, existing methods still face challenges of high construction cost, co…https://arxiv.org/abs/2611.07393https://arxiv.org/abs/2611.07393Thu, 05 Nov 2026 00:00:00 +0000Vihan Lakshman, Peter Baile Chen, Zhuohan Gu, Michael Cafarella · CIKM 2026 Workshop Recent work in machine learning has demonstrated that transformer-based language model agents can communicate state not only through text, but also directly in latent space via key-value (KV) caches. In this paper, we argue that such cache-to-cache translation can provide utility beyond the original motivating applicat…https://arxiv.org/abs/2611.06498https://arxiv.org/abs/2611.06498Wed, 04 Nov 2026 00:00:00 +0000Georgios Feretzakis, Alexandros Papaspyridis Offline retrieval-augmented generation (RAG) on edge devices is updated through signed knowledge packs that carry text and precomputed embedding vectors. A signature proves who published a pack, not that its vectors encode its text. We study this text-vector integrity gap. A verifier that recomputes vectors and accepts…https://arxiv.org/abs/2611.06170https://arxiv.org/abs/2611.06170Wed, 04 Nov 2026 00:00:00 +0000Pragyan Phukan While Large Language Models (LLMs) excel in high-resource languages, their performance often degrades in low-resource languages due to tokenization inefficiencies, generative instability, and linguistic interference. This paper presents AssameseLM-E2B and its instruction-tuned variant, AssameseLM-E2B-it, a 2-billion-pa…https://arxiv.org/abs/2611.02387https://arxiv.org/abs/2611.02387Mon, 02 Nov 2026 00:00:00 +0000Yibo Wang, Jisheng Dang, Bimei Wang, Jiongpei Hou, et al. Retrieval-augmented generation (RAG) extends large language models with external evidence, yet relevant evidence does not ensure that a generator can use it. Indiscriminate retrieval can degrade task performance while adding retrieval and long-context inference costs. Existing methods neither determine before large-sca…https://arxiv.org/abs/2610.42068https://arxiv.org/abs/2610.42068Fri, 30 Oct 2026 00:00:00 +0000Xin Ming Li, Peng Hu Low-Earth orbit (LEO) satellite networks present a challenging routing environment due to rapidly changing topology, traffic, and link conditions. Existing deep reinforcement learning (DRL) routing methods typically rely on current observations and information encoded in network weights, limiting their ability to explo…https://arxiv.org/abs/2610.39671https://arxiv.org/abs/2610.39671Thu, 29 Oct 2026 00:00:00 +0000Lingkai Kong, Yuzhe Gu, Haiteng Zhao, Zijian Wu, et al. Large language models (LLMs) draw on parametric knowledge and retrieval-augmented generation (RAG) to solve knowledge-intensive tasks. Yet problem solving in unfamiliar domains calls for tightly coupled knowledge acquisition and reasoning, beyond factual retrieval and aggregation or learning solely from a preassembled …https://arxiv.org/abs/2610.37826https://arxiv.org/abs/2610.37826Tue, 27 Oct 2026 00:00:00 +0000Dingyi Kang, Bingzhe Li Retrieval-augmented generation (RAG) lets a large language model (LLM) answer questions by retrieving relevant passages from a large corpus. However, RAG struggles with multi-hop questions, which require chaining facts spread across several documents through intermediate entities that the question never mentions. Graph…https://arxiv.org/abs/2610.35542https://arxiv.org/abs/2610.35542Mon, 26 Oct 2026 00:00:00 +0000Wenxin Zhao, Jiachuan Wang, Xinyi Zhu, Jiake Ge, et al. Long-context multimodal document question answering (QA) has become an important yet challenging task for modern vision-language models (VLMs), as it requires reasoning over evidence dispersed across pages and modalities within limited input budgets. Current pipelines adopt retrieval-augmented generation (RAG) to selec…https://arxiv.org/abs/2610.34816https://arxiv.org/abs/2610.34816Mon, 26 Oct 2026 00:00:00 +0000Qiaomu Li, Ying Xie Retrieval-augmented generation (RAG) systems often use a dedicated reranker to choose which retrieved candidates reach the language model. Reranking improves relevance but adds latency, so teams look for faster alternatives. System-1style decision models make direct, structured predictions instead of generating long re…https://arxiv.org/abs/2610.31880https://arxiv.org/abs/2610.31880Thu, 22 Oct 2026 00:00:00 +0000Meng Liang, Fucheng Wang, Lang Zhao, Jun Chen, et al. The Yangtze River Basin is a critical region for freshwater biodiversity conservation. The implementation of the Yangtze River Ten-Year Fishing Ban has increased the importance of effective aquatic biodiversity monitoring and species-level identification. However, fine-grained identification of closely related freshwat…https://arxiv.org/abs/2610.30710https://arxiv.org/abs/2610.30710Thu, 22 Oct 2026 00:00:00 +0000Guangcan Li, Shuyuan Zheng, Jiayao Liu, Shigeto Seno, et al. Over-representation analysis (ORA) is widely used to interpret gene-expression studies, but its outputs are often fragmented across gene-set databases and presented as flat term lists that obscure gene-level links among related enrichment signals. We propose Graph-Based Over-Representation Retrieval-Augmented Generatio…https://arxiv.org/abs/2610.28364https://arxiv.org/abs/2610.28364Tue, 20 Oct 2026 00:00:00 +0000Abhishek Rai Local Retrieval-Augmented Generation (RAG) systems are vulnerable to indirect prompt injection and retrieval poisoning, where malicious instructions in retrieved content are mistaken for trusted input. We present nokast-secureRAG, a roughly 0.5B-parameter small language model that acts as a semantic firewall between re…https://arxiv.org/abs/2610.26339https://arxiv.org/abs/2610.26339Mon, 19 Oct 2026 00:00:00 +0000Adrián Sánchez-Freire, Gonzalo Pérez-Gómez, Aitziber Atutxa, Ander Barrena We present a controlled, large-scale evaluation of how corpus scale, document representation, and reranking affect the effectiveness and efficiency of biomedical Retrieval-Augmented Generation (RAG) systems. Retrieval collections were progressively expanded while retaining the relevant source documents, enabling consis…https://arxiv.org/abs/2610.26951https://arxiv.org/abs/2610.26951Sun, 18 Oct 2026 00:00:00 +0000Debopriyo Banerjee, Owais Aijaz, Ahmer Tabassum, Zhuohan Xie, et al. · EMNLP 2026 Retrieval-augmented generation (RAG) is the dominant paradigm for grounding large language models in external corpora, but its similarity-driven, chunked pipeline has structural blind spots: it cannot anticipate which passages a question requires and discards document-level context that resolves cross-passage reference…https://arxiv.org/abs/2610.26938https://arxiv.org/abs/2610.26938Sat, 17 Oct 2026 00:00:00 +0000Hardik Ruparel, Raj Kumar Pandey, Pramod Kumar · SIGIR Retrieval-Augmented Generation (RAG) is increasingly served across geo-distributed cloud and edge infrastructures, where wide-area latency, cross-region egress cost, and data-residency rules compete with retrieval quality. Conventional pipelines rank by semantic similarity alone and ignore where content lives, inflatin…https://arxiv.org/abs/2610.15921https://arxiv.org/abs/2610.15921Sun, 11 Oct 2026 00:00:00 +0000Xingke Yang, Sicong Li, Lingfeng Yao, Rui Chen, et al. With ever increasing personalized large language model (LLM) services, the next question to answer is how to improve the quality of services/users' experiences, which is indicated by personalized LLM generation accuracy/quality and inference latency. Although state-of-the-art designs, e.g., parameter-efficient fine-tun…https://arxiv.org/abs/2610.15298https://arxiv.org/abs/2610.15298Sun, 11 Oct 2026 00:00:00 +0000Xiangqing Lu, Tao Zhu, Mingxing Nie, Liming Chen Large language models (LLMs) offer a new approach to wearable-sensor-based human activity recognition. Retrieval-augmented generation (RAG) further introduces similar motion examples and discriminative sensor knowledge, enabling decisions grounded in observations relevant to the current query. More evidence, however, d…https://arxiv.org/abs/2610.15186https://arxiv.org/abs/2610.15186Sat, 10 Oct 2026 00:00:00 +0000Xingjian Zhang Retrieval-augmented generation (RAG) reuses the same documents across many requests, making non-prefix KV-cache reuse attractive. Yet a cached document state is not reusable text: it is a context-conditioned causal state whose keys and values depend on the prefix, position, and preceding documents under which it was co…https://arxiv.org/abs/2610.14428https://arxiv.org/abs/2610.14428Sat, 10 Oct 2026 00:00:00 +0000Istok Bako, Bonan Min, Tiffany Shao, Claire Cheng, et al. · EMNLP 2026 Enterprise workflow automation typically requires retrieving a complete and accurate set of relevant information from large operational document corpora (SOPs, policy documents) given a specific scenario or request. In customer support (CS), for example, human specialists must determine the correct sequence of resoluti…https://arxiv.org/abs/2610.11727https://arxiv.org/abs/2610.11727Thu, 08 Oct 2026 00:00:00 +0000Yuxuan Chen, Mingwei Liu, Guangsheng Ou, Zekai Zhang, et al. Retrieval-Augmented Generation (RAG) has become a cornerstone in software engineering for enhancing Large Language Models (LLMs) with domain-specific knowledge. However, adapting retrievers to evolving code repositories remains challenging due to the noise and redundancy inherent in massive code corpora. Standard fine-…https://arxiv.org/abs/2610.08463https://arxiv.org/abs/2610.08463Tue, 06 Oct 2026 00:00:00 +0000Edan Kinderman, Elad Hoffer, Yochai Blau, Brian Chmiel, et al. Long-context inference and Retrieval-Augmented Generation (RAG) handle evidence selection at vastly different scales, from a single long prompt to an entire corpus. We ask whether a single model-internal mechanism can select evidence across this range. We introduce UNifying REtrieval And Long-Context with a Single Mode…https://arxiv.org/abs/2610.03948https://arxiv.org/abs/2610.03948Fri, 02 Oct 2026 00:00:00 +0000Xiaojun Bi, Xiaoyuan Ma, Yiwen Sun, Tianren Huang, et al. Autonomous driving decision systems must balance safety, efficiency, and social norms in complex traffic interactions. Philosophical and ethical considerations have received limited attention in existing autonomous driving decision-making approaches based on numerical optimization, sequence prediction, and large langua…https://arxiv.org/abs/2610.01767https://arxiv.org/abs/2610.01767Thu, 01 Oct 2026 00:00:00 +0000Gianluca Bonifazi, Christopher Buratti, Michele Marchetti, Federica Parlapiano, et al. Retrieval-Augmented Generation (RAG) systems for multi-hop Question Answering (QA) must balance retrieval quality with computational cost. This cost is incurred during indexing time, through the use of expensive Knowledge Graphs (KGs) or Large Language Models (LLMs) to generate summaries, or during querying, through it…https://arxiv.org/abs/2610.00923https://arxiv.org/abs/2610.00923Thu, 01 Oct 2026 00:00:00 +0000Hyojeong Yun, Jueun Kim, Wook-Shin Han Multimodal RAG retrieves text, tables, images, and videos, but choosing a retrieval granularity does not determine how much context to retain within each item. Coarse units include irrelevant content, while uniformly fine selection can remove context needed to interpret the evidence. Existing compressors address this t…https://arxiv.org/abs/2609.37426https://arxiv.org/abs/2609.37426Tue, 29 Sep 2026 00:00:00 +0000Arka Mukherjee, Kaleen Shrestha, Larissa Zhu, Maja Matarić Modern vision-language models (VLMs) have shown promising results in long-video understanding due to the rich semantic information they can capture. However, most methods focus on coarse captioning of extracted image frames that are computationally inefficient and require models with large context windows. While past w…https://arxiv.org/abs/2609.36326https://arxiv.org/abs/2609.36326Mon, 28 Sep 2026 00:00:00 +0000Truong Son Nguyen, Daniel Blackley, Ni Trieu, Evgenios M. Kornaropoulos Retrieval-augmented generation (RAG) hands the user's query to whoever hosts the corpus. We propose PILLAR, a Privacy-Preserving RAG (PPRAG) system based on Private Information Retrieval (PIR) in which a client utilizes the k documents most similar to their query from a server-held and publicly known corpus to respond …https://arxiv.org/abs/2609.35139https://arxiv.org/abs/2609.35139Mon, 28 Sep 2026 00:00:00 +0000Genglin Wang, Wangsong Yin, Yeerzhati Abudunuer, Haoxuan Xu, et al. Multi-document retrieval-augmented generation (RAG) requires a language model to process multiple retrieved text chunks before answering a question. Precomputing each chunk's KV cache independently and concatenating the caches when the chunks are retrieved can accelerate this step. However, the assembled cache lacks cr…https://arxiv.org/abs/2609.34179https://arxiv.org/abs/2609.34179Mon, 28 Sep 2026 00:00:00 +0000Richard Krueger, Lucas Krause, Zach Pocquette Retrieval-augmented generation systems are extensively instrumented with metrics, benchmarks, traces, and automated judges, but these tools do not decide whether a proposed policy change is safe to release. We present RAGWarrant, an open-source promotion-control framework that treats deployment as a constrained evidenc…https://arxiv.org/abs/2609.33503https://arxiv.org/abs/2609.33503Sun, 27 Sep 2026 00:00:00 +0000Ruoling Qi, Yirui Liu, Xuaner Wu, Yuxin Jin, et al. Cross-request KV caching reduces the prefill cost of Retrieval-Augmented Generation (RAG), but conventional prefix caching severely limits cache reuse across requests. Position-Independent Caching (PIC) removes this constraint by reusing independent chunks, but their KV states miss cross-chunk interactions. Existing me…https://arxiv.org/abs/2609.33351https://arxiv.org/abs/2609.33351Sun, 27 Sep 2026 00:00:00 +0000Hyojun Ahn, Emily Jimin Roh, Soohyun Park, Walid Saad, et al. Fidelity-based quantum retrieval ranks candidates by the fidelity between query and archive states. Applying a shared input-independent unitary after fixed state encoding leaves that fidelity unchanged, so training the circuit cannot alter the ranking. Quantum parameter-efficient input-dependent retrieval adaptation (Q…https://arxiv.org/abs/2609.29282https://arxiv.org/abs/2609.29282Thu, 24 Sep 2026 00:00:00 +0000Qi Sun, Yijia Zhang, Caibo Li, Qiang Li, et al. While graph-based and hypergraph-based Retrieval-Augmented Generation (RAG) significantly mitigate hallucinations in Large Language Models (LLMs), existing structure-based RAG systems typically adopt static traversal strategies regardless of the query complexity. We identify this ``static retrieval fallacy'' as a prima…https://arxiv.org/abs/2609.35831https://arxiv.org/abs/2609.35831Wed, 23 Sep 2026 00:00:00 +0000Isaac Olufadewa, Miracle Adesina, Ezekiel Oladejo, Owen Adeniyi, et al. Modern large language models now support context windows of more than one million tokens, which has raised the question of whether retrieval-augmented generation (RAG) is still necessary. Pure long-context (LC) processing is expensive and is known to under-attend to information placed in the middle of long inputs, whil…https://arxiv.org/abs/2609.24346https://arxiv.org/abs/2609.24346Mon, 21 Sep 2026 00:00:00 +0000Senlei Zhang, Linhao Luo, Qian-Wen Zhang, Siyu An, et al. Graph Retrieval-Augmented Generation (GraphRAG) has remarkably enhanced large language models on complex reasoning by leveraging structured entity topologies. However, existing frameworks heavily rely on standard autoregressive language models where the nature of inherent sequential generation severely hinders overall …https://arxiv.org/abs/2609.24177https://arxiv.org/abs/2609.24177Mon, 21 Sep 2026 00:00:00 +0000MD. Nafis Kamal, Mahadi Hasan Fahim, Talha Ridwan, Nadifa Zaman, et al. Legal information in Bangladesh is inaccessible to most citizens. Statutory text is English-only, trained lawyers are concentrated in urban centres, and cloud-dependent AI fails where mobile connectivity is unreliable, a setting in which hallucinated legal text causes direct harm. The system addresses statutory interpr…https://arxiv.org/abs/2609.22486https://arxiv.org/abs/2609.22486Fri, 18 Sep 2026 00:00:00 +0000Peichun Hua, Yunming Xiao Large language models (LLMs) increasingly ground their answers in proprietary data and live web content through both single-shot retrieval-augmented generation (RAG) and multi-turn agentic search. Yet the web path they use was built for human readers: search services return URLs and snippets ranked by generic relevance…https://arxiv.org/abs/2609.17983https://arxiv.org/abs/2609.17983Wed, 16 Sep 2026 00:00:00 +0000Mingyang Mao, Wyatt Mackey, Xiaomin Lin KV-cache reuse can reduce inference cost in retrieval-augmented generation and agentic systems, but cached contexts may become stale when retrieved knowledge, working memory, or user state is edited. Under causal self-attention, even a local edit can affect downstream KV states. A full re-prefill reliably restores cons…https://arxiv.org/abs/2609.17709https://arxiv.org/abs/2609.17709Tue, 15 Sep 2026 00:00:00 +0000Neeraj Anand, Payel Santra, Partha Basuchowdhuri, Debasis Ganguly, et al. Retrieval-Augmented Generation (RAG) systems typically employ fixed retriever and generator configurations across queries, despite substantial differences in query complexity and information needs, leading to inefficient allocation of computational resources. While retrieval and generation adaptivity have been studied …https://arxiv.org/abs/2609.14762https://arxiv.org/abs/2609.14762Sun, 13 Sep 2026 00:00:00 +0000Rohit Patel, Susil Kumar Mohanty, Jeenal Chaudhary Operational logs create a need for private, resource-efficient incident analysis, but aggregate detection scores can conceal severe prediction bias. We present TriCalRAG, a reproducible benchmark for log-anomaly detection with generated root-cause and remediation outputs across BGL, HDFS, Thunderbird, and OpenStack. Th…