Story
arxiv_llm_reliability ยท Nov 17, 2026 ยท paper
Source brief
StitchRAG: Constructing Compact and Evidence-Preserving Contexts for RAG with Diffusion-Based Stitching
arxiv.orgNov 17, 2026
original source linked
In brief
Retrieval-augmented generation (RAG) improves factuality by grounding large language models (LLMs) in external evidence. However, long retrieved contexts incur computational overhead and can even hurt performance due...
Feed lens
eval