Imagen de Portada

Tensor Product Representation Probes Reveal Shared Structure Across Linear Directions

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Lee, Andrew, Viégas, Fernanda, Wattenberg, Martin
Formato:	Preprint
Publicado:	2026
Materias:	Machine Learning
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

Shared Global and Local Geometry of Language Model Embeddings
por: Lee, Andrew, et al.
Publicado: (2025)

Decomposing Query-Key Feature Interactions Using Contrastive Covariances
por: Lee, Andrew, et al.
Publicado: (2026)

Relational Composition in Neural Networks: A Survey and Call to Action
por: Wattenberg, Martin, et al.
Publicado: (2024)

Dialogue Action Tokens: Steering Language Models in Goal-Directed Dialogue with a Multi-Turn Planner
por: Li, Kenneth, et al.
Publicado: (2024)

The Geometry of Self-Verification in a Task-Specific Reasoning Model
por: Lee, Andrew, et al.
Publicado: (2025)

Why Can't Transformers Learn Multiplication? Reverse-Engineering Reveals Long-Range Dependency Pitfalls
por: Bai, Xiaoyan, et al.
Publicado: (2025)

When Bad Data Leads to Good Models
por: Li, Kenneth, et al.
Publicado: (2025)

Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task
por: Li, Kenneth, et al.
Publicado: (2022)

Inference-Time Intervention: Eliciting Truthful Answers from a Language Model
por: Li, Kenneth, et al.
Publicado: (2023)

Chronotome: Real-Time Topic Modeling for Streaming Embedding Spaces
por: Lim, Matte, et al.
Publicado: (2025)

Measuring and Controlling Instruction (In)Stability in Language Model Dialogs
por: Li, Kenneth, et al.
Publicado: (2024)

Q-Probe: A Lightweight Approach to Reward Maximization for Language Models
por: Li, Kenneth, et al.
Publicado: (2024)

Story Ribbons: Reimagining Storyline Visualizations with Large Language Models
por: Yeh, Catherine, et al.
Publicado: (2025)

Optimal Representation Size: High-Dimensional Analysis of Pretraining and Linear Probing
por: Njaradi, Valentina, et al.
Publicado: (2026)

Attention-based Iterative Decomposition for Tensor Product Representation
por: Park, Taewon, et al.
Publicado: (2024)

ICLR: In-Context Learning of Representations
por: Park, Core Francisco, et al.
Publicado: (2024)

Learning Shared Representations for Multi-Task Linear Bandits
por: Lin, Jiabin, et al.
Publicado: (2026)

Manifold Steering Reveals the Shared Geometry of Neural Network Representation and Behavior
por: Wurgaft, Daniel, et al.
Publicado: (2026)

Graph Edge Representation via Tensor Product Graph Convolutional Representation
por: Jiang, Bo, et al.
Publicado: (2024)

What Does it Mean for a Neural Network to Learn a "World Model"?
por: Li, Kenneth, et al.
Publicado: (2025)

On the Linear Speedup of Personalized Federated Reinforcement Learning with Shared Representations
por: Xiong, Guojun, et al.
Publicado: (2024)

Towards Sampling Data Structures for Tensor Products in Turnstile Streams
por: Song, Zhao, et al.
Publicado: (2025)

Semantic Convergence: Investigating Shared Representations Across Scaled LLMs
por: Son, Daniel, et al.
Publicado: (2025)

Does visualization help AI understand data?
por: Li, Victoria R., et al.
Publicado: (2025)

Representation Alignment Rests on Linear Structure
por: Bangachev, Kiril, et al.
Publicado: (2026)

Gradient-Direction Sensitivity Reveals Linear-Centroid Coupling Hidden by Optimizer Trajectories
por: Xu, Yongzhong
Publicado: (2026)

E2Former: An Efficient and Equivariant Transformer with Linear-Scaling Tensor Products
por: Li, Yunyang, et al.
Publicado: (2025)

Contrastive Learning of Shared Spatiotemporal EEG Representations Across Individuals for Naturalistic Neuroscience
por: Shen, Xinke, et al.
Publicado: (2024)

Tensor Completion with Nearly Linear Samples Given Weak Side Information
por: Yu, Christina Lee, et al.
Publicado: (2020)

Rhetorical Questions in LLM Representations: A Linear Probing Study
por: Yao, Louie Hong, et al.
Publicado: (2026)

A Single Direction of Truth: An Observer Model's Linear Residual Probe Exposes and Steers Contextual Hallucinations
por: O'Neill, Charles, et al.
Publicado: (2025)

Fully Distributed, Flexible Compositional Visual Representations via Soft Tensor Products
por: Sun, Bethia, et al.
Publicado: (2024)

Learning Fine-grained Parameter Sharing via Sparse Tensor Decomposition
por: Üyük, Cem, et al.
Publicado: (2024)

Structured Unitary Tensor Network Representations for Circuit-Efficient Quantum Data Encoding
por: Lin, Guang, et al.
Publicado: (2026)

Near-optimal and Efficient First-Order Algorithm for Multi-Task Learning with Shared Linear Representation
por: Ding, Shihong, et al.
Publicado: (2026)

High-Dimensional Tensor Discriminant Analysis: Low-Rank Discriminant Structure, Representation Synergy, and Theoretical Guarantees
por: Chen, Elynn, et al.
Publicado: (2025)

Learning Representations for Reasoning: Generalizing Across Diverse Structures
por: Zhu, Zhaocheng
Publicado: (2024)

Dynamics Reveals Structure: Challenging the Linear Propagation Assumption
por: Chang, Hoyeon, et al.
Publicado: (2026)

SharedRep-RLHF: A Shared Representation Approach to RLHF with Diverse Preferences
por: Mukherjee, Arpan, et al.
Publicado: (2025)

Shared Parameter Subspaces and Cross-Task Linearity in Emergently Misaligned Behavior
por: Arturi, Daniel Aarao Reis, et al.
Publicado: (2025)