Latest research and updates
Research papers, new tools, best practices, and industry developments in the RAG and generative AI ecosystem.
Microsoft Research unveils GraphRAG, a novel approach that combines RAG with knowledge graphs to improve contextual understanding
Recent research reveals new document chunking approaches that significantly improve RAG system performance
UC Berkeley researchers introduce DecomposeRAG, an automated query decomposition framework that significantly improves multi-hop question answering.
Anthropic releases Claude 3.5 Sonnet with extended context window, improved citation accuracy, and new RAG-specific features for enterprise applications.
New GPT-4.5 Turbo model features built-in retrieval capabilities, structured output mode, and 50% cost reduction for RAG applications.
Cohere's new embedding model delivers state-of-the-art performance on MTEB benchmark while reducing dimensions from 1024 to 768, cutting costs and improving speed.
Google Research introduces AutoRAGEval, an automated evaluation framework that reliably assesses RAG quality without human annotation.
Weaviate's new hybrid search engine combines BM25, vector search, and learned ranking in a single optimized index for superior RAG retrieval.
MIT study demonstrates that two-stage retrieval with cross-encoder reranking significantly outperforms single-stage vector search across multiple benchmarks.
Stanford and DeepMind researchers present MM-RAG, a unified framework for retrieving and reasoning over multiple modalities with 65% accuracy improvement.
Updated BEIR benchmark includes 6 new datasets, adversarial examples, and improved evaluation methodology for more robust retrieval testing.
Microsoft Research unveils GraphRAG 2.0, featuring improved entity extraction, relationship mapping, and 40% better accuracy on complex multi-hop queries.