AI Model Analysis Comparison Hub
4 papers - avg viability 2.8
Top Papers
- From Atoms to Trees: Building a Structured Feature Forest with Hierarchical Sparse Autoencoders(4.0)
Introducing HSAE, a scalable tool for building and analyzing hierarchical conceptual structures in LLMs using sparse autoencoders.
- SYNAPSE: Framework for Neuron Analysis and Perturbation in Sequence Encoding(3.0)
Develop a lightweight framework, SYNAPSE, for stress-testing and understanding Transformer models' robustness without retraining.
- Do Reasoning Models Enhance Embedding Models?(2.0)
Exploration into how reasoning models influence embedding models reveals no clear performance advantage.
- Cross-Architecture Model Diffing with Crosscoders: Unsupervised Discovery of Differences Between LLMs(2.0)
This paper explores cross-architecture model diffing to identify behavioral differences in AI models.