Weights Read and Write Features: Scalable Parameter Decomposition Grounded in Activation Space
cs.LG
Submitted: 2026-09-29
Updated: 2026-10-01
Code: https://github.com/tue147/weights-read-write
Sources
- Eliciting Latent Predictions from Transformers with the Tuned Lens
- Interpretability in Parameter Space: Minimizing Mechanistic Description Length with Attribution-based Parameter Decomposition
- Stochastic Parameter Decomposition
- BatchTopK Sparse Autoencoders
- Learning Multi-Level Features with Matryoshka Sparse Autoencoders
- Semantic Optimal Transport for Sparse Autoencoder Feature Matching and Circuit Compression
- Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders
- A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders
- Low-Rank Adapting Models for Sparse Autoencoders
- Identifying Sparsely Active Circuits Through Local Loss Landscape Decomposition
- The Pile: An 800GB Dataset of Diverse Text for Language Modeling
- Scaling and evaluating sparse autoencoders
- Patchscopes: A Unifying Framework for Inspecting Hidden Representations of Language Models
- Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
- Analyze Feature Flow to Enhance Interpretation and Steering in Language Models
- Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2
- Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models
- Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning
- The Linear Representation Hypothesis and the Geometry of Large Language Models
- Evaluating SAE interpretability without explanations
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks