NAVER LABS re-implements pipeline for IWSLT 2026 instruction-following

Paper describes re-implementation of NAVER LABS' IWSLT 2025 pipeline for the 2026 constrained short audio track, using SeamlessM4T-v2-large as speech encoder. A separate NAVER LABS Europe submission jointly performs ASR, ST, and SQA.