Linear probes for Othello board states factor into tensor-product structure with square and color embeddings composed by a binding matrix, from which the linear probes can be directly recovered.
Emergent linear representations in world models of self-supervised sequence models
10 Pith papers cite this work. Polarity classification is still indexing.
10
Pith papers citing it
representative citing papers
Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.
A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.
LLMs perform in-context learning as trajectories through a structured low-dimensional conceptual belief space, with the structure visible in both behavior and internal representations and causally manipulable via interventions.
Tool identity is linearly readable and steerable in LLMs via mean activation differences, with 77-100% switch accuracy and error prediction from activation gaps.
The Linear Accessibility Profile predicts steering vector effectiveness and optimal layers with Spearman correlations of 0.86-0.91 using unembedding projections on intermediate states across multiple models and concepts.
Multi-turn neural transparency using behavioral vectors and dynamic visualizations improves user anticipation and evaluation of LLM trait expression while reducing overconfidence, per a randomized study with 246 participants.
Overthinking in medical QA is linearly decodable at 71.6% accuracy yet fixed residual-stream steering yields no correction across 29 configurations, while enabling selective abstention with AUROC 0.610.
Varying evaluation metrics and corruption methods in activation patching produces different localization and circuit discovery outcomes in language models, leading to recommendations for preferred practices.
GeoMathCode interleaves math reasoning with programmatic code outputs for geometry problems in MLLMs and shows that reasoning steps and hierarchical code structures become disentangled in latent space after SFT.
citing papers explorer
-
Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions
Linear probes for Othello board states factor into tensor-product structure with square and color embeddings composed by a binding matrix, from which the linear probes can be directly recovered.
-
Refusal in Language Models Is Mediated by a Single Direction
Refusal in language models is mediated by a single direction in residual stream activations that can be erased to disable safety or added to elicit refusal.
-
Correcting Influence: Unboxing LLM Outputs with Orthogonal Latent Spaces
A latent mediation framework with sparse autoencoders enables non-additive token-level influence attribution in LLMs by learning orthogonal features and back-propagating attributions.
-
Stories in Space: In-Context Learning Trajectories in Conceptual Belief Space
LLMs perform in-context learning as trajectories through a structured low-dimensional conceptual belief space, with the structure visible in both behavior and internal representations and causally manipulable via interventions.
-
Tool Calling is Linearly Readable and Steerable in Language Models
Tool identity is linearly readable and steerable in LLMs via mean activation differences, with 77-100% switch accuracy and error prediction from activation gaps.
-
Predicting Where Steering Vectors Succeed
The Linear Accessibility Profile predicts steering vector effectiveness and optimal layers with Spearman correlations of 0.86-0.91 using unembedding projections on intermediate states across multiple models and concepts.
-
Multi-Turn Neural Transparency: Surfacing Neural Activations Improves User Calibration to LLM Behavioral Drift
Multi-turn neural transparency using behavioral vectors and dynamic visualizations improves user anticipation and evaluation of LLM trait expression while reducing overconfidence, per a randomized study with 246 participants.
-
Decodable but Not Corrected by Fixed Residual-Stream Linear Steering: Evidence from Medical LLM Failure Regimes
Overthinking in medical QA is linearly decodable at 71.6% accuracy yet fixed residual-stream steering yields no correction across 29 configurations, while enabling selective abstention with AUROC 0.610.
-
Towards Best Practices of Activation Patching in Language Models: Metrics and Methods
Varying evaluation metrics and corruption methods in activation patching produces different localization and circuit discovery outcomes in language models, leading to recommendations for preferred practices.
-
GeoMathCode: Understanding Interleaved Math-Code Reasoning for Geometry Problem Solving
GeoMathCode interleaves math reasoning with programmatic code outputs for geometry problems in MLLMs and shows that reasoning steps and hierarchical code structures become disentangled in latent space after SFT.