Representational Atoms in Neural Networks
cs.CL, cs.AI
Submitted: 2026-11-10
Updated: 2026-11-10
Code: https://github.com/thunlp/OpenHowNet
Terminology
Sources
- Toy Models of Superposition
- Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoders
- OpenHowNet: An Open Sememe-based Lexical Knowledge Base
- Llama Scope: Extracting Millions of Features from Llama-3.1-8B with Sparse Autoencoders
- The Linear Representation Hypothesis and the Geometry of Large Language Models
- The Restricted Isometry of ReLU Networks: Generalization through Norm Concentration
- Conditioning of Random Feature Matrices: Double Descent and Generalization Error
- Dataless Model Selection with the Deep Frame Potential
- Compute Optimal Inference and Provable Amortisation Gap in Sparse Autoencoders
- Improving Dictionary Learning with Gated Sparse Autoencoders
- Gemma 2: Improving Open Language Models at a Practical Size
- The Llama 3 Herd of Models
- Qwen3.5-Omni Technical Report
- Olmo Hybrid: From Theory to Practice and Back
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data