Search papers, labs, and topics across Lattice.
To address data redundancy in terminology-aware translation, the authors prune glossary-annotated training sets to retain exclusively "hard" instances where the unconstrained model output directly contradicts the prescribed glossary. In controlled evaluations at fixed data volume, this selective filtering boosts term accuracy from 78.7% to 89.9%, allowing simple supervised fine-tuning to surpass prior reinforcement learning (GRPO) baselines. Deployed as SalamandraTA-7b-instruct v3.0 within a document-level inference pipeline, the approach attained 94.2% term success at 74.6 chrF++ in the WMT 2026 Terminology Shared Task.
Discarding "easy" examples where the base model already follows glossary constraints yields an 11-point accuracy leap at fixed data volume, outperforming RL-based alignment methods with standard supervised fine-tuning alone.
Terminology-aware translation asks for more than a correct translation: the output must use the exact terms a glossary prescribes. The standard recipe, fine-tuning on glossary-annotated translation pairs, hides an inefficiency: for most examples the glossary prescribes exactly what the model would have produced anyway, so they teach nothing about following a glossary. We therefore keep only the examples where the model's own translation contradicts the glossary. In a controlled study at fixed data volume, this selection alone raises term accuracy from 78.7% to 89.9%. The filtered data, built by a two-way synthetic pipeline on open models, is part of the instruction-tuning mixture of our public release SalamandraTA-7b-instruct v3.0, which, used exactly as released and wrapped in a document-level inference pipeline, forms the BSC submission to the WMT26 Terminology Shared Task Track 1. At the official WMT26 evaluation, our system achieves 94.2% term success at 74.6 chrF++, with only two of the twenty-two submissions outperforming it on both metrics. On last year's benchmark, it also surpasses our GRPO-based system, despite being trained solely with ordinary supervised fine-tuning.
