Author

Dim P. Papadopoulos

1 paper indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Preprint Oct 2026

SIMIT: Self-Improving Vision-Language Models via Imagination at Test-Time

SIMIT is introduced, a test-time self-improvement framework in which a vision-language model creates its own query-specific training data before answering an unlabeled test query to improve performance on a given target test set without annotations or external models.

M. Kaya, Desmond Elliott, Dim P. Papadopoulos · 0 citations