LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs

By Ziyan Jiang, Xue-Guang Ma, Wenhu Chen. 2024

Subfields: Retrieval-Augmented Generation, Large Language Models, Question Answering & Reading Comprehension

Citations in the corpus

• Context is Gold to find the Gold Passage: Evaluating and Training Contextual Document Embeddings (2025, unknown): “…et al., 2025) along with long context embedding models (Zhu et al., 2024), modern document retrieval pipelines typically segment lengthy documents into smaller chunks to optimize the granularity for efficient retrieval and readability of the retrieved content (Xu et al., 2024; Jiang et al., 2024).”
• ReasonIR: Training Retrievers for Reasoning Tasks (2025, unknown): “For LLM reranker baselines, we compare with RankLLaMA-7B (Ma et al., 2024) and Rank1-7B (Weller et al., 2025), which rerank the top-100 documents retrieved by BM25 using GPT-4 R EASON - QUERY .”
• Length-Induced Embedding Collapse in PLM-based Models (2024, unknown): “We conduct experiments using three commonly used long-text datasets, including nq and hotpotqa in LongRAG (Jiang et al., 2024) and LongAlpaca-12k (Chen et al., 2024b), which are selected based on relevance to the keyword "long" and chosen for their evenly distributed text lengths.”