The field of Retrieval-Augmented Generation (RAG) is rapidly evolving, with a focus on improving the efficiency and effectiveness of Large Language Models (LLMs) in generating high-quality text. Recent developments have centered on addressing the challenges of knowledge updates, hallucination issues, and the need for more reliable and trustworthy AI-generated content. Researchers have proposed various innovative approaches, including the integration of polyviews, graph-based knowledge integration, and hybrid model collaboration. Noteworthy papers in this area include AlignRAG, which introduces a novel framework for resolving misalignments in RAG pipelines, and PolyRAG, which incorporates judges from different perspectives to improve retrieval-augmented generation in medical applications. The use of LLMs as data annotators has also shown promise, with studies demonstrating their potential to automate the annotation process and improve the quality of training data. Furthermore, the development of new evaluation metrics and benchmarks, such as MIRAGE, has enabled more accurate assessments of RAG systems. Overall, the field of RAG is advancing rapidly, with a growing emphasis on developing more reliable, efficient, and effective systems for real-world applications.
Advancements in Retrieval-Augmented Generation
Sources
RAG Without the Lag: Interactive Debugging for Retrieval-Augmented Generation Pipelines
Long-context Non-factoid Question Answering in Indic Languages
Transparentize the Internal and External Knowledge Utilization in LLMs with Trustworthy Citation
AlignRAG: An Adaptable Framework for Resolving Misalignments in Retrieval-Aware Reasoning of RAG
Retrieval Augmented Generation Evaluation in the Era of Large Language Models: A Comprehensive Survey
ColBERT-serve: Efficient Multi-Stage Memory-Mapped Scoring
POLYRAG: Integrating Polyviews into Retrieval-Augmented Generation for Medical Applications
Vector Embedding, Retrieval-Augmented Generation, CPU-NPU Collaboration, Heterogeneous Computing
LLMs as Data Annotators: How Close Are We to Human Performance
The Great Nugget Recall: Automating Fact Extraction and RAG Evaluation with Large Language Models
Support Evaluation for the TREC 2024 RAG Track: Comparing Human versus LLM Judges
RAGDoll: Efficient Offloading-based Online RAG System on a Single GPU
CiteFix: Enhancing RAG Accuracy Through Post-Processing Citation Correction
The Viability of Crowdsourcing for RAG Evaluation
Synergizing RAG and Reasoning: A Systematic Review
ConTextual: Improving Clinical Text Summarization in LLMs with Context-preserving Token Filtering and Knowledge Graphs
A Unified Retrieval Framework with Document Ranking and EDU Filtering for Multi-document Summarization
Credible plan-driven RAG method for Multi-hop Question Answering
LLM-assisted Graph-RAG Information Extraction from IFC Data
MIRAGE: A Metric-Intensive Benchmark for Retrieval-Augmented Generation Evaluation
A RAG-Based Multi-Agent LLM System for Natural Hazard Resilience and Adaptation
Adaptive Orchestration of Modular Generative Information Access Systems
Replication and Exploration of Generative Retrieval over Dynamic Corpora
CLIPSE -- a minimalistic CLIP-based image search engine for research