arXiv AI

Position: Use Sparse Autoencoders to Discover Unknowns

arXiv:2506. 23845v2 Announce Type: replace-cross Abstract: While sparse autoencoders (SAEs) have generated significant excitement, a series of negative results have added to skepticism about their usefulness.

arXiv AI
4d ago

Making Clinical Language Models Auditable: Concept-Guided Fine-Tuning for Robust Prediction

The paper introduces CAST, a concept-guided artifact suppression tuning framework that uses sparse autoencoders to identify and suppress note-specific artifacts in clinical language models. CAST labels latent features with an LLM-assisted pipeline and ICD‑10 constraints, then fine‑tunes the model while providing post‑hoc per‑concept attributions for auditability. In experiments on MIMIC‑IV discharge‑note mortality prediction, CAST outperforms standard fine‑tuned encoders and competes with strong LLM baselines while offering a feature‑level audit trail of clinical concepts and suppressed artifacts.

By Jin Mu, Guanhua Chen
arXiv Machine Learning
Jun 18

From Sparse Features to Trustworthy Proxies: Certifying SAE-Based Interpretability

arXiv:2606. 18383v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) are increasingly used to extract interpretable features from language models (LMs), yet a central question remains: when can an SAE-based explanation be treated as a faithful view of an underlying frozen LM We study this through a post-hoc generalization framework that certifies the LM via a sparse proxy, obtained by replacing a native hidden activation with its pretrained SAE reconstruction.

By Dibyanayan Bandyopadhyay, Asif Ekbal
Hugging Face Trending Papers
Aug 13

SAEVerbalizer: Generating Explanations for Sparse Autoencoder Features via Representation Verbalization

Sparse autoencoders (SAEs) are proposed to extract numerous features from large language model (LLM) representations, yet explaining these features still relies primarily on external observation. This reliance leads to superficial explanations inferred from observed model behavior and computational inefficiency from collecting such behavioral evidence at scale.

arXiv Machine Learning
5d ago

ICON Decomposition: Multivariate Concept-Level Explanations of Deep Representations for Model Auditing

ICON Decomposition is a new method for explaining deep neural networks by quantifying how much variance each concept explains in a network layer after accounting for all other concepts and the outcome. Unlike previous concept‑based methods that evaluate concepts in isolation, ICON can distinguish genuine model reliance from spurious correlations. Experiments on synthetic data, skin‑lesion, and brain‑imaging models show that ICON recovers concept importance more accurately, isolates truly relied‑upon concepts, and provides sparse explanations validated through retraining and out‑of‑distribution testing.

By Roshan Prakash Rane, Marco Simnacher, Manuel Pfeuffer, Marc-Andre Schulz, Nys Tjade Siegel, Maximilian Dreyer, Frederik Pahde, Wojciech Samek, Sonja Greven, Kerstin Ritter
arXiv Machine Learning
Jul 14

TimeSAE: Causal Sparse Decoding for Faithful Explanations of Black-Box Time Series Models

arXiv:2601. 09776v2 Announce Type: replace Abstract: As black box models and pretrained models gain traction in time series applications, understanding and explaining their predictions becomes increasingly vital, especially in high-stakes domains where interpretability and trust are essential.

By Khalid Oublal, Quentin Bouniot, Qi Gan, Stephan Cl\'emen\c{c}on, Zeynep Akata