LLM Digest

Story

arxiv_llm_reliability · Sep 1, 2026 · paper

Source brief

SCoNE: Selective Context-aware Neuron Editing for Robust Retrieval-Augmented Generation

arxiv.orgSep 1, 2026
original source linked

In brief

Retrieval-Augmented Generation (RAG) is highly sensitive to retrieval noise: when retrieved documents mix informative and irrelevant context, LLMs are easily distracted, leading to hallucinations. To overcome this, we...

Feed lens

eval

Read the original at arxiv.org →Open in live feed Read that day’s brief

SCoNE: Selective Context-aware Neuron Editing for Robust Retrieval-Augmented Generation

Earlier in this thread 4 items

WattCouncil: Context-Aware Household Energy Scenario Generation With Governed LLMs

Cognitive-structured Multimodal Agent for Multimodal Understanding, Generation, and Editing

Medi-Gemma: A Hybrid Clinical Decision Support System Integrating Deterministic EMR Analytics and Retrieval-Augmented Generation

Efficient Retrieval-Augmented Generation via Token Co-occurrence Graphs