Fingerprinting Inference Systems of Large Language Models

Anna Wimbauer, Jonas Möller, Erik Imgrund, Konrad Rieck · 2026

Read Fingerprinting Inference Systems of Large Language Models by Anna Wimbauer, Jonas Möller, Erik Imgrund, Konrad Rieck (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-29979.txt.

Description

Anna Wimbauer, Jonas Möller, Erik Imgrund, and Konrad Rieck show model-running machinery leaves a trace in a language model's words. Engines, attention back ends, and hardware bend numbers, so the same model answers a shade differently per setup. Those quirks reach visible text, open to any querier. Their fingerprint method reads prompt-response behavior and names engine, attention back end, and hardware platform even at non-zero temperature. Blocking such tracing looks fundamentally hard, since it would need identical numbers across all stacks, so partial defenses are offered.

Catalog record

CREATORAnna Wimbauer, Jonas Möller, Erik Imgrund, Konrad Rieck
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39455
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.29979

Browse the full library · About the Tower of Babel