AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit

Nicholas Jiang, Xiaoqing Sun, Lisa Dunlap, Lewis Smith, Neel Nanda

ICML 2026 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM representations · 0 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 SAE embeddings instead of LLM-based summarize/cluster baselines for dataset diffingSAE embeddings · Model-output diffing (six comparisons)average verified frequency difference per hypothesisImproved
use SAEs instead of pure LLMs for hypothesis generationSAE embeddings · Dataset diffingtoken cost: 2-8x lower costImproved
use SAE embeddings instead of an LLM for finding injected correlationsSAE embeddings · Synthetic correlations in Pile (10 shuffles)recovery of synthetic correlations: SAE found all 4; LLM found 1/10 to 9/10Improved
use SAE method instead of LLM and correlated topic model baselines on real dataSAE embeddings · CivilComments and Pileverified NPMI of discovered pairsImproved
filter SAE latents to reasoning-related ones for targeted clusteringSAE embeddings · GSM8K solutionsclustering along reasoning styleImproved
use SAE embeddings instead of dense embeddings on property-based retrievalSAE embeddings · Six-dataset property retrieval benchmarkMAPImproved
skip LLM reranking of retrieved latents (ablation)SAE retrieval · Property-based retrievalMAPHurt
relabel latents with the Pile and LMSYS-1M (ablation)SAE retrieval · Property-based retrievalMAPMixed
combine strongest baseline (OpenAI+LLM) with SAE scoresSAE retrieval · Property-based retrievalMAPImproved
increase reader model size from 8B to 70B (ablation)SAE latents · GPT-5 responses (chat-like data)median F1 of latentsImproved
increase reader model size from 8B to 70B (ablation)SAE latents · arXiv and PileF1 of latentsNo clear effect
use SAE retrieval on data far from the SAE training distributionSAE retrieval · Biology abstracts and short storiesMAPMixed

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