Imagen de Portada

How to use and interpret activation patching

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Heimersheim, Stefan, Nanda, Neel
Formato:	Preprint
Publicado:	2024
Materias:	Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

You can remove GPT2's LayerNorm by fine-tuning
por: Heimersheim, Stefan
Publicado: (2024)

Towards eliciting latent knowledge from LLMs with mechanistic interpretability
por: Cywiński, Bartosz, et al.
Publicado: (2025)

Investigating Sensitive Directions in GPT-2: An Improved Baseline and Comparative Analysis of SAEs
por: Lee, Daniel J., et al.
Publicado: (2024)

Explorations of Self-Repair in Language Models
por: Rushing, Cody, et al.
Publicado: (2024)

Towards Best Practices of Activation Patching in Language Models: Metrics and Methods
por: Zhang, Fred, et al.
Publicado: (2023)

Towards Principled Evaluations of Sparse Autoencoders for Interpretability and Control
por: Makelov, Aleksandar, et al.
Publicado: (2024)

Difficulties with Evaluating a Deception Detector for AIs
por: Smith, Lewis, et al.
Publicado: (2025)

Base Models Know How to Reason, Thinking Models Learn When
por: Venhoff, Constantin, et al.
Publicado: (2025)

Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models
por: Leask, Patrick, et al.
Publicado: (2025)

BatchTopK Sparse Autoencoders
por: Bussmann, Bart, et al.
Publicado: (2024)

Transcoders Find Interpretable LLM Feature Circuits
por: Dunefsky, Jacob, et al.
Publicado: (2024)

Summing Up the Facts: Additive Mechanisms Behind Factual Recall in LLMs
por: Chughtai, Bilal, et al.
Publicado: (2024)

Detecting Strategic Deception Using Linear Probes
por: Goldowsky-Dill, Nicholas, et al.
Publicado: (2025)

How Visual Representations Map to Language Feature Space in Multimodal LLMs
por: Venhoff, Constantin, et al.
Publicado: (2025)

Reasoning-Finetuning Repurposes Latent Representations in Base Models
por: Ward, Jake, et al.
Publicado: (2025)

The Obfuscation Atlas: Mapping Where Honesty Emerges in RLVR with Deception Probes
por: Taufeeque, Mohammad, et al.
Publicado: (2026)

Evolution of SAE Features Across Layers in LLMs
por: Balcells, Daniel, et al.
Publicado: (2024)

Transformers Don't Need LayerNorm at Inference Time: Scaling LayerNorm Removal to GPT-2 XL and the Implications for Mechanistic Interpretability
por: Baroni, Luca, et al.
Publicado: (2025)

Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition
por: Braun, Dan, et al.
Publicado: (2025)

RelP: Faithful and Efficient Circuit Discovery in Language Models via Relevance Patching
por: Jafari, Farnoush Rezaei, et al.
Publicado: (2025)

Evaluating Sparse Autoencoders on Targeted Concept Erasure Tasks
por: Karvonen, Adam, et al.
Publicado: (2024)

AtP*: An efficient and scalable method for localizing LLM behaviour to components
por: Kramár, János, et al.
Publicado: (2024)

Learning Multi-Level Features with Matryoshka Sparse Autoencoders
por: Bussmann, Bart, et al.
Publicado: (2025)

Convergent Linear Representations of Emergent Misalignment
por: Soligo, Anna, et al.
Publicado: (2025)

Evaluating Synthetic Activations composed of SAE Latents in GPT-2
por: Giglemiani, Giorgi, et al.
Publicado: (2024)

Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrieval
por: Venhoff, Constantin, et al.
Publicado: (2025)

Do I Know This Entity? Knowledge Awareness and Hallucinations in Language Models
por: Ferrando, Javier, et al.
Publicado: (2024)

Interpreting Attention Layer Outputs with Sparse Autoencoders
por: Kissane, Connor, et al.
Publicado: (2024)

Benchmarking Deception Probes via Black-to-White Performance Boosts
por: Parrack, Avi, et al.
Publicado: (2025)

Characterizing stable regions in the residual stream of LLMs
por: Janiak, Jett, et al.
Publicado: (2024)

SCALAR: Benchmarking SAE Interaction Sparsity in Toy LLMs
por: Fillingham, Sean P., et al.
Publicado: (2025)

Model Organisms for Emergent Misalignment
por: Turner, Edward, et al.
Publicado: (2025)

Understanding Reasoning in Thinking Language Models via Steering Vectors
por: Venhoff, Constantin, et al.
Publicado: (2025)

Are Sparse Autoencoders Useful? A Case Study in Sparse Probing
por: Kantamneni, Subhash, et al.
Publicado: (2025)

Interpretable Embeddings with Sparse Autoencoders: A Data Analysis Toolkit
por: Jiang, Nick, et al.
Publicado: (2025)

Simple LLM Baselines are Competitive for Model Diffing
por: Kempf, Elias, et al.
Publicado: (2026)

Thought Branches: Interpreting LLM Reasoning Requires Resampling
por: Macar, Uzay, et al.
Publicado: (2025)

Thought Anchors: Which LLM Reasoning Steps Matter?
por: Bogdan, Paul C., et al.
Publicado: (2025)

Simple Mechanistic Explanations for Out-Of-Context Reasoning
por: Wang, Atticus, et al.
Publicado: (2025)

What's the plan? Metrics for implicit planning in LLMs and their application to rhyme generation and question answering
por: Maar, Jim, et al.
Publicado: (2026)