BSC_IWSLT26_IF_SHORT_unconstrained_primary

Data conditions: unconstrained

Model architecture: direct (SeamlessM4Tv2 speech encoder and SalamandraTA-7B-wmt25 as the LLM backbone)

Track: Short