← Search

Dániel Varga

2 accepted papers

2025

The Structure of Relation Decoding Linear Operators in Large Language Models

NeurIPS 2025spotlight

This paper investigates the structure of linear operators introduced in Hernandez et al. [2023] that decode specific relational facts in transformer language models. We extend their single-relation findings to a collection of relations and systematically chart their organization. We show that such c…

Cited by 0SourceScholar
2021

Similarity and Matching of Neural Network Representations

NeurIPS 2021poster

We employ a toolset --- dubbed Dr. Frankenstein --- to analyse the similarity of representations in deep neural networks. With this toolset we aim to match the activations on given layers of two trained neural networks by joining them with a stitching layer. We demonstrate that the inner representat…