4 papers
Log-Likelihood, Simpson's Paradox, and the Detection of Machine-Generated Text
Tom Kempton, Viktor Drobnyi, Maeve Madigan +1
The ability to reliably distinguish human-written text from that generated by large language models is of profound societal importance. The dominant approach to this problem exploi…
Local dimension spectrum for dominated planar self-affine sets
Alex Batsis, Antti Käenmäki, Tom Kempton
The local dimension spectrum provides a framework for quantifying the fractal properties of a measure, and it is well understood for non-overlapping self-similar measures. In this…
Local Normalization Distortion and the Thermodynamic Formalism of Decoding Strategies for Large Language Models
Tom Kempton, Stuart Burrell
Advances in hardware and language model architecture have spurred a revolution in natural language generation. However, autoregressive models compute probability distributions over…
TempTest: Local Normalization Distortion and the Detection of Machine-generated Text
Tom Kempton, Stuart Burrell, Connor Cheverall
Existing methods for the zero-shot detection of machine-generated text are dominated by three statistical quantities: log-likelihood, log-rank, and entropy. As language models mimi…