AI Paper Atlas

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

Interpretable and Steerable Concept Bottleneck Sparse Autoencoders

Akshay Kulkarni, Tsui-Wei Weng, Vivek Narayanaswamy, Shusen Liu, Wesam A. Sakla, Kowshik Thopalli

CVPR 2026 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM representations · 6 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
prune low-utility SAE neurons and add concept bottleneck (CB-SAE)SAE · LLaVA-1.5-7BCLIP-Dissect interpretability: 0.154 -> 0.244Improved
CB-SAESAE · LLaVA-1.5-7Bsteerability (unit vector): 0.198 -> 0.261Improved
CB-SAESAE · LLaVA-MOREsteerability (unit vector): 0.179 -> 0.192Improved
CB-SAESAE · LLaVA-MORECLIP-Dissect interpretability: 0.194 -> 0.291Improved
CB-SAESAE · UnCLIPsteerability (white image): 0.654 -> 0.664No clear effect
CB-SAESAE · UnCLIPmonosemanticity: 0.540 -> 0.594Improved
add steerability cyclic loss Lst (ablation)CB-SAE · LLaVA-1.5-7Bsteerability: +2.9% steerabilityImproved
prune SAE with interpretability+steerability sum instead of one score (ablation)CB-SAE · LLaVA-1.5-7Binterpretability and steerabilityMixed
retain fewer SAE neurons after pruning (ablation)CB-SAE · LLaVA-1.5-7Binterpretability and steerabilityMixed
use CB neurons vs retained SAE neurons (ablation)CB-SAE · LLaVA-1.5-7Binterpretability: 0.238 -> 0.323Improved
use CB neurons vs retained SAE neurons (ablation)CB-SAE · LLaVA-1.5-7Bsteerability (unit vector): 0.263 -> 0.231Hurt
keep discarded SAE neurons (ablation)SAE · LLaVA-1.5-7BCLIP-Dissect interpretability: 0.238 (retained) -> 0.084 (discarded)Hurt

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