retrieval-augmented generation

Existing document retrieval methods suffer from two key limitations: scalar similarity scores lack the expressiveness for fine-grained semantic understanding, and overreliance on global representations neglects local semantic alignment between queries and documents. To address these issues, we propose GeAR (Generative-enhanced Retrieval), a novel framework that integrates a lightweight decoder into the dual-encoder retrieval paradigm—enabling fine-grained query-document alignment and interpretable natural language generation without increasing inference latency. GeAR jointly leverages cross-modal representations and employs large language models to synthesize high-quality training data. Evaluated across multiple benchmark datasets, GeAR achieves state-of-the-art performance in both retrieval accuracy and passage localization, while additionally generating human-readable supporting evidence. The framework thus advances the trade-off among efficiency, effectiveness, and interpretability in neural retrieval.

This study investigates whether retrieval fusion techniques—commonly adopted in real-world retrieval-augmented generation (RAG) systems, such as multi-query retrieval and reciprocal rank fusion—consistently improve end-to-end answer quality under practical deployment constraints. Conducted within an enterprise knowledge-base RAG pipeline, the evaluation is performed under fixed retrieval depth, reranking budget, and latency limits. While retrieval fusion enhances initial recall, it fails to translate into improved Top-k accuracy after subsequent reranking and context truncation; notably, Hit@10 declines from 0.51 to 0.48 and incurs additional latency. These findings challenge the prevailing assumption of the default efficacy of recall-oriented fusion strategies, revealing diminishing returns in production settings where downstream processing and system constraints critically shape overall performance.

This study systematically investigates the trade-off between retrieval accuracy and generation flexibility in knowledge-intensive tasks—namely, open-domain question answering, document re-ranking, and retrieval-augmented language modeling. Within a unified evaluation framework, it conducts the first cross-model comparison of representative retrieval-based (BM25), generative (GPT-4-o), and hybrid models (e.g., RAG) across three benchmarks: Natural Questions (NQ), BEIR, and WikiText-103, using standardized metrics including nDCG@10 and top-1 accuracy. Results show that DPR achieves 50.17% top-1 accuracy on NQ; hybrid RAG improves BEIR’s average nDCG@10 from 43.42 to 52.59; and BM25 yields the lowest perplexity on WikiText-103. Based on these findings, the study proposes empirically grounded model selection criteria aligned with task-specific semantic requirements and factual constraint characteristics, offering both methodological guidance and practical evidence for robust RAG system design.

This study addresses the challenge of retrieving scientific literature implicitly referenced in social media posts. We propose a three-stage hybrid retrieval framework: (1) a dual-path dense retrieval layer combining BM25 and a fine-tuned INF-Retriever-v1; (2) an efficient vector indexing layer built upon FAISS; and (3) a re-ranking layer using an open-source LLM-based cross-encoder. The entire pipeline runs locally without external training data, ensuring reproducibility and practicality. To our knowledge, this is the first work to synergistically integrate lightweight dense retrieval with large language model–based re-ranking for semantic alignment between informal social media language and formal academic texts, effectively bridging their lexical and conceptual gaps. On the official benchmark, our approach achieves 76.46% MRR@5 on the development set (rank #1) and 66.43% on the blind test set (rank #3 among 31 teams), approaching the performance of the top-performing system.

This work addresses the challenge of deploying a shared retrieval backbone in industrial systems, where balancing performance and deployment flexibility across multiple downstream tasks remains difficult. To overcome the limitations of conventional approaches that rely on a single optimal checkpoint, the authors propose a multi-stage optimization framework that tailors component-level and hybrid-stage configuration strategies to the distinct performance characteristics of dense retrievers and rerankers throughout training. This approach significantly enhances the adaptability of the shared backbone and improves overall retrieval effectiveness. End-to-end evaluation demonstrates that the resulting shared retrieval service has been successfully deployed across multiple industrial applications, delivering substantial gains in both system performance and scalability.

This work addresses the inefficiencies of conventional text chunking in standard retrieval-augmented generation (RAG) systems, which often introduces redundancy, leading to excessive storage costs and degraded retrieval performance. To mitigate this, the authors propose a lightweight pre-index filtering mechanism that integrates semantic similarity, topic coherence, and named entity recognition to selectively prune redundant text chunks prior to indexing. Evaluated through a token-level precision, recall, and intersection-over-union framework, the approach reduces vector index size by 25%–36% while preserving retrieval quality comparable to that of the original system, thereby significantly enhancing the overall efficiency of RAG pipelines.

This study investigates optimal text chunking strategies for enhancing the response quality of Retrieval-Augmented Generation (RAG) systems when applied to structurally complex academic papers. We systematically compare semantic clustering, fixed-length, and recursive chunking approaches, evaluating output faithfulness and relevance using the RAGAs framework. To our knowledge, this is the first empirical comparison of multiple chunking strategies on long-form scholarly texts. Our findings indicate that semantic clustering does not significantly outperform simpler methods, and that question type—generic versus document-specific—substantially influences system performance. Furthermore, we identify limitations in the reliability of RAGAs’ faithfulness metric for such tasks, suggesting a need for more robust evaluation measures in academic RAG applications.

This study addresses a critical oversight in current retrieval-augmented generation (RAG) systems: their reliance on human-oriented document representations, which neglect the distinct representational needs of large language models as content consumers. Under fixed retrieval results, the authors systematically evaluate the impact of 14 document representation strategies—including selection, summarization, and rewriting—on question-answering accuracy across four generative models. Introducing answer retention rate as a novel metric to assess whether transformed documents preserve the correct answer, controlled experiments reveal for the first time that answer retention is the primary driver of generation accuracy, challenging prior assumptions that attributed performance gains to specific representational mechanisms. Notably, when answer retention is high, variations in wording, structure, length, or query dependence exert minimal influence on accuracy, underscoring that preserving answer information outweighs representational form.

This work addresses key challenges in deploying large language models for retrieval-augmented generation (RAG), including high computational overhead, rapid knowledge obsolescence, and manual dependency in component selection. The authors propose a modular evaluation framework that, for the first time, directly links hardware constraints to RAG performance. By integrating resource telemetry with an automated recommendation mechanism, the framework efficiently identifies optimal combinations of components—including document chunking strategies, embedding models, vector databases, and retrievers—for domain-specific datasets. This approach maintains high generation quality while substantially reducing resource consumption. Designed to support rapid prototyping on consumer-grade hardware, the framework enables automatic, domain-tailored RAG configuration, achieving a favorable trade-off among accuracy, efficiency, and scalability.

This study addresses the lack of systematic evaluation of modern retrieval methods in Retrieval-Augmented Generation (RAG) systems for heterogeneous financial documents containing both text and tables. It presents the first large-scale benchmark comparing ten retrieval strategies—including sparse, dense, hybrid fusion, cross-encoder reranking, and query expansion—on a financial mixed-document question answering task. Evaluation employs multiple metrics (Recall@k, MRR, nDCG, and numerical match accuracy) alongside bootstrap significance testing. Results reveal that BM25 outperforms state-of-the-art dense retrievers in this domain; a two-stage hybrid retrieval pipeline with neural reranking achieves substantial gains (Recall@5 = 0.816, MRR@3 = 0.605); context-aware retrieval consistently improves performance, whereas query expansion yields limited benefits for numerical questions. The authors release a comprehensive open-source benchmark and provide practical guidance on cost–accuracy trade-offs.