ACADATA: Parallel Dataset of Academic Data for Machine Translation

TL;DR

ACAData introduces a high-quality multilingual parallel corpus for academic translation, comprising ACAD-train (~1.5M paragraph pairs across 12 languages and 96 directions) and ACAD-bench (~5,944 test translations across 12 directions). By fine-tuning Salamandra LLMs on ACAD-train, the study demonstrates consistent gains in academic translation quality (up to +12.4 d-BLEU for 2B models) and substantial improvements in long-context translation on a general-domain benchmark (up to 24.9% win-margin when translating out of English). ACAD-bench proves challenging for non-domain models, with the fine-tuned models outperforming several proprietary systems on academic translation while providing a robust, publicly available benchmark for long-context translation research. The work also provides a detailed data construction pipeline (OAI-PMH harvesting, LaBSE-SLIDE similarity, GlotLID filtering, normalization, and deduplication) and releases all resources under permissive licenses to accelerate research in domain-specific MT and long-context modeling.

Abstract

We present ACADATA, a high-quality parallel dataset for academic translation, that consists of two subsets: ACAD-TRAIN, which contains approximately 1.5 million author-generated paragraph pairs across 96 language directions and ACAD-BENCH, a curated evaluation set of almost 6,000 translations covering 12 directions. To validate its utility, we fine-tune two Large Language Models (LLMs) on ACAD-TRAIN and benchmark them on ACAD-BENCH against specialized machine-translation systems, general-purpose, open-weight LLMs, and several large-scale proprietary models. Experimental results demonstrate that fine-tuning on ACAD-TRAIN leads to improvements in academic translation quality by +6.1 and +12.4 d-BLEU points on average for 7B and 2B models respectively, while also improving long-context translation in a general domain by up to 24.9% when translating out of English. The fine-tuned top-performing model surpasses the best propietary and open-weight models on academic translation domain. By releasing ACAD-TRAIN, ACAD-BENCH and the fine-tuned models, we provide the community with a valuable resource to advance research in academic domain and long-context translation.

Figures (13)

• Figure 1: Translation quality on ACAD-bench in xx→en directions for models fine-tuned with ACAD-train and a set of open-weight and proprietary systems of different scales. When the scale is not known, we represent it with a horizontal line.
• Figure 2: Relative distribution of data across languages in ACAData. Each bar represents a source language, with segments showing the proportion of sentence pairs per target language. Percentages are normalized per language, while absolute pair counts are indicated above each bar.
• Figure 3: Win margin (%) over the base model as a function of source length, divided into quartiles ($q_1$–$q_4$) based on token count. We compare the performance of Salamandra-2B + ACAD-train and Salamandra-7B + ACAD-train against the base model across translation directions in the WMT24++ benchmark. The left plot shows results for English-to-Other (en→xx), and the right for Other-to-English (xx→en) directions.
• Figure 4: Boxplots for the 62 directions shared between WMT24++ and ACAD-train. The plots show the distribution of BLEU scores (left) and COMET scores (right). Each boxplot compares the model's performance on WMT24++ dataset before (Baseline) and after fine-tuning on the ACAD-train dataset (+ACAD-train).
• Figure 5: Bar plot showing all the domains in ACAD-bench.