Copertina

Causality is Key for Interpretability Claims to Generalise

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Joshi, Shruti, Mueller, Aaron, Klindt, David, Brendel, Wieland, Reizinger, Patrik, Sridhar, Dhanya
Natura:	Preprint
Pubblicazione:	2026
Soggetti:	Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Who Guards the Guardians? The Challenges of Evaluating Identifiability of Learned Representations
di: Joshi, Shruti, et al.
Pubblicazione: (2026)

Position: An Empirically Grounded Identifiability Theory Will Accelerate Self-Supervised Learning Research
di: Reizinger, Patrik, et al.
Pubblicazione: (2025)

From Isolation to Entanglement: When Do Interpretability Methods Identify and Disentangle Known Concepts?
di: Mueller, Aaron, et al.
Pubblicazione: (2025)

Identifiable Exchangeable Mechanisms for Causal Structure and Representation Learning
di: Reizinger, Patrik, et al.
Pubblicazione: (2024)

Estimating Treatment Effects with Independent Component Analysis
di: Reizinger, Patrik, et al.
Pubblicazione: (2025)

Cross-Entropy Is All You Need To Invert the Data Generating Process
di: Reizinger, Patrik, et al.
Pubblicazione: (2024)

An Interventional Perspective on Identifiability in Gaussian LTI Systems with Independent Component Analysis
di: Rajendran, Goutham, et al.
Pubblicazione: (2023)

Rule Extrapolation in Language Models: A Study of Compositional Generalization on OOD Prompts
di: Mészáros, Anna, et al.
Pubblicazione: (2024)

Position: Understanding LLMs Requires More Than Statistical Generalization
di: Reizinger, Patrik, et al.
Pubblicazione: (2024)

Skill Learning via Policy Diversity Yields Identifiable Representations for Reinforcement Learning
di: Reizinger, Patrik, et al.
Pubblicazione: (2025)

InfoNCE: Identifying the Gap Between Theory and Practice
di: Rusak, Evgenia, et al.
Pubblicazione: (2024)

From superposition to sparse codes: interpretable representations in neural networks
di: Klindt, David, et al.
Pubblicazione: (2025)

Stop Probing, Start Coding: Why Linear Probes and Sparse Autoencoders Fail at Compositional Generalisation
di: Pacela, Vitória Barin, et al.
Pubblicazione: (2026)

Sparse Shift Autoencoders for Identifying Concepts from Large Language Model Activations
di: Joshi, Shruti, et al.
Pubblicazione: (2025)

Out-of-distribution Tests Reveal Compositionality in Chess Transformers
di: Mészáros, Anna, et al.
Pubblicazione: (2025)

General Causal Imputation via Synthetic Interventions
di: Jiralerspong, Marco, et al.
Pubblicazione: (2024)

The Landscape of Causal Discovery Data: Grounding Causal Discovery in Real-World Applications
di: Brouillard, Philippe, et al.
Pubblicazione: (2024)

Missed Causes and Ambiguous Effects: Counterfactuals Pose Challenges for Interpreting Neural Networks
di: Mueller, Aaron
Pubblicazione: (2024)

Data Whitening Improves Sparse Autoencoder Learning
di: Saraswatula, Ashwin, et al.
Pubblicazione: (2025)

Causal Representation Learning in Temporal Data via Single-Parent Decoding
di: Brouillard, Philippe, et al.
Pubblicazione: (2024)

Pretraining Frequency Predicts Compositional Generalization of CLIP on Real-World Tasks
di: Wiedemer, Thaddäus, et al.
Pubblicazione: (2025)

MATH-Beyond: A Benchmark for RL to Expand Beyond the Base Model
di: Mayilvahanan, Prasanna, et al.
Pubblicazione: (2025)

Identifiable Deep Generative Models via Sparse Decoding
di: Moran, Gemma E., et al.
Pubblicazione: (2021)

Demystifying amortized causal discovery with transformers
di: Montagna, Francesco, et al.
Pubblicazione: (2024)

Superposition disentanglement of neural representations reveals hidden alignment
di: Longon, André, et al.
Pubblicazione: (2025)

The Role of Causal Features in Strategic Classification for Robustness and Alignment
di: Gois, Antonio, et al.
Pubblicazione: (2026)

Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
di: Marks, Samuel, et al.
Pubblicazione: (2024)

In-Context Learning Can Re-learn Forbidden Tasks
di: Xhonneux, Sophie, et al.
Pubblicazione: (2024)

Provable Compositional Generalization for Object-Centric Learning
di: Wiedemer, Thaddäus, et al.
Pubblicazione: (2023)

Sparsity regularization via tree-structured environments for disentangled representations
di: Layne, Elliot, et al.
Pubblicazione: (2024)

Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoders
di: O'Neill, Charles, et al.
Pubblicazione: (2024)

Occam's Razor for Self Supervised Learning: What is Sufficient to Learn Good Representations?
di: Ibrahim, Mark, et al.
Pubblicazione: (2024)

LAION-C: An Out-of-Distribution Benchmark for Web-Scale Vision Models
di: Li, Fanfei, et al.
Pubblicazione: (2025)

LLMs on the Line: Data Determines Loss-to-Loss Scaling Laws
di: Mayilvahanan, Prasanna, et al.
Pubblicazione: (2025)

Generation is Required for Data-Efficient Perception
di: Brady, Jack, et al.
Pubblicazione: (2025)

Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?
di: Mayilvahanan, Prasanna, et al.
Pubblicazione: (2023)

CSQL: Mapping Documents into Causal Databases
di: Mahadevan, Sridhar
Pubblicazione: (2026)

The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysis
di: Mueller, Aaron, et al.
Pubblicazione: (2024)

Stress-Testing Causal Claims via Cardinality Repairs
di: Gabbay, Yarden, et al.
Pubblicazione: (2025)

Interaction Asymmetry: A General Principle for Learning Composable Abstractions
di: Brady, Jack, et al.
Pubblicazione: (2024)