arXiv Machine Learning
Jun 15
Ensembling Sparse Autoencoders
arXiv:2505. 16077v2 Announce Type: replace Abstract: Sparse autoencoders (SAEs) are used to decompose neural network activations into human-interpretable features.
Building Auditable AI Text Detectors with Sparse Autoencoders
The paper introduces a method to transform opaque fine‑tuned neural AI‑text detectors into interpretable classifiers by training sparse autoencoders on the detector’s residual stream. Each autoencoder feature is labeled with a human‑readable concept, and only features meeting a fidelity threshold are kept. A lightweight classifier built on these interpretable features outperforms both neural and stylometric detectors, especially on out‑of‑distribution texts.
Related stories
Hugging Face Trending Papers
Aug 25
Linear Probing Provides Robust and Efficient Detection of Machine-Generated Text
Distinguishing machine-generated text (MGT) from human-written text (HWT) becomes increasingly important due to potential misuse. However, most supervised detectors often degrade out-of-domain (OOD) a...
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 AI
Aug 28
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.
arXiv Machine Learning
Jul 21
Persistent Sparse Autoencoders: Learning Feature Timescales in Language Models
arXiv:2607. 17117v1 Announce Type: new Abstract: Sparse autoencoders (SAEs) decompose language model activations into sparse features, but standard SAEs encode each token independently and do not expose information that persists across a sequence.
arXiv AI
Jun 4
A Systematic Analysis of Linguistic Features in AI-Generated Text Detection Across Domains and Models
arXiv:2606. 04177v1 Announce Type: cross Abstract: Interpretable linguistic features offer a promising approach for explaining why a given text appears machine-generated, particularly for non-expert users.
arXiv Computation and Language
3d ago
A Systematic Analysis of Linguistic Features in AI-Generated English Text Detection Across Domains and Models
The paper presents a large‑scale empirical study of 284 interpretable linguistic features to detect AI‑generated English text across 30 language models and 10 text domains. It finds that classifiers using only linguistic features can reliably distinguish machine‑from human‑written text, but most indicators are highly context‑dependent. Only lexical richness measures consistently generalize across model families and domains, offering robust signals for detection.
arXiv AI
Oct 27
Robust and Fine-Grained Detection of AI Generated Texts
arXiv:2504.11952v4 Announce Type: replace-cross Abstract: An ideal detection system for machine generated content is supposed to work well on any generator as many more advanced LLMs come into existe...
arXiv AI
Jun 6
LLM Self-Recognition: Steering and Retrieving Activation Signatures
arXiv:2606. 06315v1 Announce Type: new Abstract: Recent advances in interpretability suggest that large language models (LLMs) implicitly encode signals in their generated text that enable self-recognition of their outputs.
arXiv Computation and Language
Oct 19
Beyond Surface Cues: Robust LLM-Generated Korean Text Detection
The paper introduces “KoLing”, a bias‑controlled adversarial benchmark of 50.9k Korean instances designed to suppress superficial cues while preserving semantic content, and evaluates both LLM‑generated and LLM‑modified text detection. It also proposes “KoLingProbe”, an interpretable detector grounded in theoretical linguistics that models Korean‑specific properties across multiple linguistic levels. Experiments demonstrate that KoLingProbe outperforms strong baselines on all detection tasks and remains robust to adversarial attacks, with analyses showing that integrating complementary linguistic features across multiple levels improves both performance and interpretability.
arXiv Machine Learning
Jun 17
Stable and Steerable Sparse Autoencoders with Weight Regularization
arXiv:2603. 04198v2 Announce Type: replace-cross Abstract: Sparse autoencoders (SAEs) are widely used to extract human-interpretable features from neural network activations, but their learned features can vary substantially across random seeds and training choices.
arXiv Machine Learning
Jun 2
AdaptiveK: Complexity-Driven Sparse Autoencoders for Interpretable Language Model Representations
arXiv:2508. 17320v3 Announce Type: replace Abstract: Understanding the internal representations of large language models (LLMs) remains a central challenge for interpretability research.