AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Toward Faithful Retrieval-Augmented Generation with Sparse Autoencoders

Guangzhi Xiong, Zhenghao He, Bohan Liu, Sanchit Sinha, Aidong Zhang

ICLR 2026 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use RAGLens (SAE features + MI selection + GAM) instead of best internal-representation baseline ReDeEPLlama2-7B · RAGTruthAUC: 0.7458 -> 0.8413Improved
use RAGLens instead of ReDeEPLlama2-13B · Dolly (Accurate Context)AUC: 0.8420 -> 0.8568Improved
use pre-activation instead of post-activation SAE features (ablation)RAGLens · TofuEvalaccuracy: 0.5637 -> 0.6865Improved
use Transcoder instead of SAE as feature extractor (ablation)RAGLens · RAGTruthaccuracy: 0.7810 -> 0.7778No clear effect
train detector on RAGTruth, test zero-shot on AggreFact vs CoT promptingRAGLens · AggreFactAUROC: 0.5741 -> 0.8019Improved
train on AggreFact and test on RAGTruthRAGLens · RAGTruthAUROC: 0.4842 -> 0.5330No clear effect
train on Data2txt, test on SummaryRAGLens · RAGTruth SummaryAUROC: 0.4924 -> 0.5386Mixed
add instance-level feedback from RAGLens for revisionLlama2-7B · RAGTruth mitigation (human)hallucination rate: 71.11% -> 62.22%Improved
use token-level instead of instance-level feedback (ablation)Llama2-7B · RAGTruth mitigation (human)hallucination rate: 62.22% -> 55.56%Improved
use MI-based feature selection instead of random selection (ablation)RAGLens · RAGTruth, Dollydetection performanceImproved
use GAM predictor instead of LR, MLP or XGBoost (ablation)RAGLens · multiple datasetsdetection performanceImproved
use SAE features instead of raw hidden states at small K' (ablation)RAGLens · AggreFact, TofuEvaldetection performanceImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy