The field of Retrieval-Augmented Generation (RAG) is rapidly evolving, with a focus on improving the accuracy and reliability of large language models by incorporating external knowledge into their input prompts. Recent developments have seen the introduction of novel frameworks and techniques, such as Role-Augmented Intent-Driven Generative Search Engine Optimization and Geo-RAG, which aim to enhance the capabilities of RAG in various domains, including geoscience and legal research. Noteworthy papers in this area include 'Role-Augmented Intent-Driven Generative Search Engine Optimization', which proposes a structured optimization pathway for generative search engines, and 'RAG for Geoscience', which envisions a next-generation paradigm for geoscience workflows. These advancements have the potential to significantly impact the field, enabling more trustworthy and transparent workflows, and improving the overall performance of RAG systems.
Advancements in Retrieval-Augmented Generation
Sources
Role-Augmented Intent-Driven Generative Search Engine Optimization
RAG for Geoscience: What We Expect, Gaps and Opportunities
Retrieval-augmented reasoning with lean language models
HumorPlanSearch: Structured Planning and HuCoT for Contextual AI Humor
GridCodex: A RAG-Driven AI Framework for Power Grid Code Reasoning and Compliance
All for law and law for all: Adaptive RAG Pipeline for Legal Research
FLAIR: Feedback Learning for Adaptive Information Retrieval
Revisiting RAG Ensemble: A Theoretical and Mechanistic Analysis of Multi-RAG System Collaboration
Ask Good Questions for Large Language Models
An automatic patent literature retrieval system based on LLM-RAG
Retrieval-Augmented Generation in Industry: An Interview Study on Use Cases, Requirements, Challenges, and Evaluation
Retrieval-Augmented Review Generation for Poisoning Recommender Systems
Conflict-Aware Soft Prompting for Retrieval-Augmented Generation
Adversarial Attacks against Neural Ranking Models via In-Context Learning
DesignCLIP: Multimodal Learning with CLIP for Design Patent Understanding
Test-time Corpus Feedback: From Retrieval to RAG