arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~

AI 大模型

RAG / 检索增强生成

检索增强生成、向量检索、知识库问答和面向大模型的搜索系统。

2026-04-10 至 2026-04-10 共收录 5 信号源:cs.IR, cs.CL, cs.AI, cs.DB

1. RAG评测 5 篇

2604.07964 2026-04-10 cs.AI cs.LG 79%

Are we still able to recognize pearls? Machine-driven peer review and the risk to creativity: An explainable RAG-XAI detection framework with markers extraction

我们仍然能识别珍珠吗?机器驱动的同行评审与创造力风险:一个可解释的RAG-XAI检测框架与标记提取

Alin-Gabriel Văduva, Simona-Vasilica Oprea, Adela Bâra

机构 * Bucharest University of Economic Studies(布加勒斯特经济研究大学)

专题命中 RAG评测 :RAG(title,abstract);分类 cs.AI

AI总结 本文提出RAG-XAI框架,通过标记提取检测自动化评审,以保留科学的透明度和创造力,实验显示其在检测性能上显著优于传统方法。

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.18019 2026-04-10 cs.CL cs.AI cs.SE 62%

BenchBrowser: Retrieving Evidence for Evaluating Benchmark Validity

BenchBrowser:用于评估基准有效性证据的检索

Harshita Diddee, Gregory Yauney, Swabha Swayamdipta, Daphne Ippolito

机构 * Language Technologies Institute, Carnegie Mellon University(卡内基梅隆大学语言技术研究所) Thomas Lord Department of Computer Science, University of Southern California(南加州大学托马斯·洛德计算机科学系)

专题命中 RAG评测 :retriever(abstract);分类 cs.CL、cs.AI

AI总结 BenchBrowser通过检索20个基准测试集中的相关评估项目,帮助评估者诊断基准测试在内容有效性与收敛有效性上的不足,从而弥补实践者意图与实际测试内容之间的差距。

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.07428 2026-04-10 cs.LG cs.AI 57%

Regret-Aware Policy Optimization: Environment-Level Memory for Replay Suppression under Delayed Harm

考虑后悔的策略优化:环境层面的记忆用于回放抑制在延迟伤害下

Prakul Sunil Hiremath

机构 * Department of Computer Science and Engineering, VTU, Belagavi, India(印度贝尔高姆VTU计算机科学与工程系) Aliens on Earth (AoE) Autonomous Research Group, Belagavi, India(印度贝尔高姆地球外星人自主研究小组)

专题命中 RAG评测 :RAG(abstract);分类 cs.AI

AI总结 本文提出RAPO方法,通过环境层面的记忆机制抑制回放,减少有害区域的可达性,从而提升安全性和任务性能。

Comments 18 pages, 3 figures. Includes theoretical analysis and experiments on graph diffusion environments

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.14997 2026-04-10 cs.CL cs.AI cs.IR 56%

OrgForge: A Multi-Agent Simulation Framework for Verifiable Synthetic Corporate Corpora

OrgForge:一种用于可验证合成企业语料的多智能体仿真框架

Jeffrey Flynt

专题命中 RAG评测 :分类 cs.IR、cs.CL、cs.AI;RAG(comments)

AI总结 OrgForge通过多智能体仿真框架生成一致且可追溯的合成企业语料,解决现有语料在法律约束和幻觉问题上的不足,提升AI系统评估的准确性。

Comments v2: Major revision. Recenters the paper on the simulation framework as the primary contribution. System Architecture substantially expanded (CRM state machine, Knowledge Recovery Arc, multi-pathway knowledge gap detection, embedding-based ticket assignment). Introduction restructured for broader framing. RAG retrieval baselines replaced by cross-document consistency evaluation

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.08082 2026-04-10 cs.HC 50%

From Binary Groundedness to Support Relations: Towards a Reader-Centred Taxonomy for Comprehension of AI Output

从二元 groundedness 到支持关系:迈向以读者为中心的 AI 输出理解分类体系

Advait Sarkar, Christian Poelitz, Viktor Kewenig

专题命中 RAG评测 :retrieval augmented generation(abstract)

AI总结 本文提出以读者为中心的 groundedness 分类体系,旨在通过支持关系提升 AI 输出理解的透明度和可解释性。

Comments Advait Sarkar, Christian Poelitz, and Viktor Kewenig. 2026. From Binary Groundedness to Support Relations: Towards a Reader-Centred Taxonomy for Comprehension of AI Output. ACM CHI 2026 Workshop on Science and Technology for Augmenting Reading (CHI '26 STAR) ACM CHI 2026 Workshop on Science and Technology for Augmenting Reading (CHI '26 STAR)

详情

展开后加载摘要…

URL PDF HTML 收藏