One-shot Style Transfer LLM log-probabilities for Authorship Attribution and Verification
This work introduces an unsupervised framework that uses the log-probabilities of an LLM to measure style transferability between two texts, and proposes an additional mechanism that increases test-time computation to improve accuracy; enabling flexible trade-offs between computational cost and task performance.
Pablo Miralles-González, Javier Huertas-Tato, Alejandro Martín et al.
· 3 citations