Cover Image

Position: Mechanistic Interpretability Should Prioritize Feature Consistency in SAEs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Song, Xiangchen, Muhamed, Aashiq, Zheng, Yujia, Kong, Lingjing, Tang, Zeyu, Diab, Mona T., Smith, Virginia, Zhang, Kun
Format:	Preprint
Published:	2025
Subjects:	Machine Learning Artificial Intelligence Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

SAEs $\textit{Can}$ Improve Unlearning: Dynamic Sparse Autoencoder Guardrails for Precision Unlearning in LLMs
by: Muhamed, Aashiq, et al.
Published: (2025)

Decoding Dark Matter: Specialized Sparse Autoencoders for Interpreting Rare Concepts in Foundation Models
by: Muhamed, Aashiq, et al.
Published: (2024)

CoRAG: Collaborative Retrieval-Augmented Generation
by: Muhamed, Aashiq, et al.
Published: (2025)

DSPA: Dynamic SAE Steering for Data-Efficient Preference Alignment
by: Wedgwood, James, et al.
Published: (2026)

RefusalBench: Generative Evaluation of Selective Refusal in Grounded Language Models
by: Muhamed, Aashiq, et al.
Published: (2025)

Pando: Do Interpretability Methods Work When Models Won't Explain Themselves?
by: Zhong, Ziqian, et al.
Published: (2026)

CCRS: A Zero-Shot LLM-as-a-Judge Framework for Comprehensive RAG Evaluation
by: Muhamed, Aashiq
Published: (2025)

Beyond Understanding: Evaluating the Pragmatic Gap in LLMs' Cultural Processing of Figurative Language
by: Attia, Mena, et al.
Published: (2025)

Grass: Compute Efficient Low-Memory LLM Training with Structured Sparse Gradients
by: Muhamed, Aashiq, et al.
Published: (2024)

SAEs Are Good for Steering -- If You Select the Right Features
by: Arad, Dana, et al.
Published: (2025)

StressRoBERTa: Cross-Condition Transfer Learning from Depression, Anxiety, and PTSD to Stress Detection
by: Alqahtani, Amal, et al.
Published: (2025)

Towards Self-Refinement of Vision-Language Models with Triangular Consistency
by: Deng, Yunlong, et al.
Published: (2025)

Resa: Transparent Reasoning Models via SAEs
by: Wang, Shangshang, et al.
Published: (2025)

Emotion Classification in Low and Moderate Resource Languages
by: Tafreshi, Shabnam, et al.
Published: (2024)

SimBA: Simplifying Benchmark Analysis Using Performance Matrices Alone
by: Subramani, Nishant, et al.
Published: (2025)

AI Safety Should Prioritize the Future of Work
by: Hazra, Sanchaita, et al.
Published: (2025)

Mechanistic Interpretability with SAEs: Probing Religion, Violence, and Geography in Large Language Models
by: Simbeck, Katharina, et al.
Published: (2025)

Advancing Reasoning in Diffusion Language Models with Denoising Process Rewards
by: Xie, Shaoan, et al.
Published: (2025)

Teach Old SAEs New Domain Tricks with Boosting
by: Koriagin, Nikita, et al.
Published: (2025)

SmartCLIP: Modular Vision-language Alignment with Identification Guarantees
by: Xie, Shaoan, et al.
Published: (2025)

Reflection-Window Decoding: Text Generation with Selective Refinement
by: Tang, Zeyu, et al.
Published: (2025)

Personal Information Parroting in Language Models
by: Subramani, Nishant, et al.
Published: (2026)

Mechanistic Interpretability Needs Philosophy
by: Williams, Iwan, et al.
Published: (2025)

DLM-Scope: Mechanistic Interpretability of Diffusion Language Models via Sparse Autoencoders
by: Wang, Xu, et al.
Published: (2026)

Position: Pause Recycling LoRAs and Prioritize Mechanisms to Uncover Limits and Effectiveness
by: Chen, Mei-Yen, et al.
Published: (2025)

Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens
by: Deng, Yunlong, et al.
Published: (2025)

Sentipolis: Emotion-Aware Agents for Social Simulations
by: Fu, Chiyuan, et al.
Published: (2026)

Mechanistic Interpretability of Emotion Inference in Large Language Models
by: Tak, Ala N., et al.
Published: (2025)

Dissecting Bias in LLMs: A Mechanistic Interpretability Perspective
by: Chandna, Bhavik, et al.
Published: (2025)

A Note on Bias to Complete
by: Xu, Jia, et al.
Published: (2024)

MIB: A Mechanistic Interpretability Benchmark
by: Mueller, Aaron, et al.
Published: (2025)

Toward a Theory of Generalizability in LLM Mechanistic Interpretability Research
by: Trott, Sean
Published: (2025)

Analysing Moral Bias in Finetuned LLMs through Mechanistic Interpretability
by: Raimondi, Bianca, et al.
Published: (2025)

LLM Interpretability with Identifiable Temporal-Instantaneous Representation
by: Song, Xiangchen, et al.
Published: (2025)

Mechanistic Interpretability of GPT-2: Lexical and Contextual Layers in Sentiment Analysis
by: Hatua, Amartya
Published: (2025)

Towards Understanding and Improving Refusal in Compressed Models via Mechanistic Interpretability
by: Chhabra, Vishnu Kabir, et al.
Published: (2025)

Position: Mechanistic Interpretability Must Disclose Identification Assumptions for Causal Claims
by: Lin, Zezheng, et al.
Published: (2026)

Beyond the Black Box: Identifiable Interpretation and Control in Generative Models via Causal Minimality
by: Kong, Lingjing, et al.
Published: (2025)

Mechanistic Interpretability of Socio-Political Frames in Language Models
by: Asghari, Hadi, et al.
Published: (2025)

HyperDAS: Towards Automating Mechanistic Interpretability with Hypernetworks
by: Sun, Jiuding, et al.
Published: (2025)