retrieval-augmented-generation

TRACE is a training-free agentic retrieval framework designed for accountable source discovery in historical digital archives, outperforming RAG baselines on a benchmark with low cost per question.

This paper evaluates intrinsic sequence-likelihood confidence in retrieval-dominated extractive QA, finding that confidence signals are insufficient for model control and that retrieval alone achieves high accuracy under pre-specified criteria.

TrioRAG is a graph-free multimodal RAG framework that uses multi-signal late fusion for efficient cross-document question answering, matching or outperforming graph-based systems. It also introduces AutoQA, an automotive benchmark with web-sourced images.

This paper benchmarks Knowledge Graph-based augmentation against Retrieval Augmented Generation for culturally specific question answering, demonstrating significant error reduction on Latin American culture data using the LatamQA dataset.

This position paper argues that legal LLM hallucinations should be evaluated as failures of legal warrant rather than factual inaccuracies, proposing a new benchmark framework for assessing legal AI systems.

This paper proposes a Tri-Metric Router, a deterministic framework for adaptive routing among inference pipelines to address the Compression Paradox in long-context RAG on commodity GPUs, achieving zero OOM failures and improved performance.

ORDER introduces a framework that dynamically adapts indexing and retrieval strategies in RAG systems based on the query, improving performance in expert domains.

The paper proposes a query-aware triage layer for retrieval-augmented generation (RAG) pipelines to handle the material relationship of sources to queries, using scoring and routing methods evaluated on synthetic data to define risk-coverage targets.

This paper evaluates the faithfulness of AquiLLM, an open-weight RAG-LLM system, for scientific research in astronomy, showing strong performance on retrieval tasks but challenges in synthesis.

The paper proposes CLEAR, an agentic framework for cross-source evidence adjudication to improve large language models in medicine by handling conflicts from multiple knowledge sources. It demonstrates competitive performance across benchmarks, with significant gains in settings where direct inference or retrieval is weak.

RAG-CT is a novel defense method that identifies malicious queries by analyzing entropy and margin distributions to mitigate privacy risks in Retrieval-Augmented Generation systems, significantly reducing PII leakage.

This paper formalizes semantic shadowing in mutable RAG and introduces GC-Mem, a temporal dominance-based protocol that resolves conflicts and recovers over 90% accuracy.

The paper proposes a retrieval-augmented generation framework using large language models to recommend safety countermeasures from crash narratives, achieving high precision and recall in evaluation on real-world data.

ClinAgent is a ReAct-based agent system using agentic RAG to enable natural language querying of clinical trial information, with evaluation across multiple LLM backends showing complementary strengths.

This paper investigates the trade-off between context compression and citation attribution in retrieval-augmented generation, evaluating multiple compression methods and revealing significant gaps between answer quality and attribution accuracy.

The paper presents ShopEase, a generative AI-based multi-agent framework for enterprise customer support using hybrid retrieval-augmented generation. It evaluates retrieval configurations, finding dense retrieval with FAISS achieves the best accuracy.

This paper introduces the Atomic User Model (AUM), a structured representation for organizing user personality to enhance large language model interactions, showing improved personalization and context efficiency in simulations.

This paper investigates whether retrieval signals provide additional routing value beyond the query in adaptive multimodal RAG systems, concluding that they do not consistently improve decision-making.

CoG is a cognitive-inspired, training-free framework for adaptive knowledge exploration in retrieval-augmented generation, achieving state-of-the-art performance on multi-hop QA benchmarks through plan-explore-reflect cycles and bidirectional graph-text synergy.

The paper introduces EAR, an entity-aware partitioning approach for retrieval-augmented generation in multiple-choice question answering, which extracts anchors from questions and corpora to retrieve compact windows instead of fixed-size chunks.