Top RAG Papers of the Week (November Week 2, 2024)

[1] Optimal Search and Retrieval for RAG

This paper investigates how retrievers can be optimized for RAG pipelines for common tasks such as Question Answering (QA). The authors unveil a number of insights useful to practitioners developing high-performance RAG pipelines. For example, lowering search accuracy has minor implications for RAG performance while potentially increasing retrieval speed and memory efficiency. [Tweet] and [Paper]

[2] Benchmarking Library for RAG

Inconsistent RAG benchmarking makes it difficult to compare approaches and understand the impact of each component in the pipeline. To address this, the authors introduce BERGEN, a benchmarking library for RAG. This library is useful for reproducible research standardizing RAG experiments. BERGEN also supports multilingual datasets to promote RAG development beyond English. [Tweet] and [Paper]

[3] AssistRAG

This paper introduces AssistRAG, integrating an intelligent information assistant within LLMs. This assistant manages memory and knowledge through tool usage, action execution, memory building, and plan specification. AssistRAG enhances information retrieval and decision-making. Experiments show AssistRAG significantly outperforms benchmarks, especially benefiting less advanced LLMs. [Tweet] and [Paper]

"Top LLM Papers of the Week" newsletter is read by over 21k+ AI Researchers, Engineers and Developers. If you would like to promote with us, contact Kalyan KS

[4] Analyzing RAG Systems through Sufficient Context

This paper analyzes RAG systems through a new lens called “sufficient context”. Results reveal that that proprietary LLMs (Gemini, GPT, Claude) excel at answering queries when the context is sufficient, but often output incorrect answers instead of abstaining when the context is not. On the other hand, open-source LLMs (Llama, Mistral, Gemma) hallucinate or abstain often, even with sufficient context.? [Tweet] and [Paper]

[5] RAG-based Semi-Supervised Text Classification

This paper introduces a? semi-supervised learning approach involving RAG for text classification. The proposed approach integrates few-shot learning with retrieval-augmented generation (RAG) and conventional statistical clustering.? The proposed approach demonstrates SOTA results, with few-shot augmented data alone producing results nearly equivalent to those achieved with fully labeled datasets. [Tweet] and [Paper]

Do subscribe to the newsletter so that you won't miss interesting updates related to Generative AI, LLMs and RAG.

Kalyan KS, Research Scientist(NLP) at Akmmus AI Labs