Tag
This paper presents the development of parallel and monolingual corpora for scientific machine translation across Spanish-English, French-English, and Portuguese-English, targeting four domains: Cancer Research, Energy Research, Neuroscience, and Transportation. The corpora are used to fine-tune neural machine translation systems, addressing challenges of specialized vocabulary and syntax in scientific text.
University of Florida Gators submission to the AmericasNLP 2026 shared task on cultural image captioning for Indigenous languages, using a two-stage pipeline with Qwen2.5-VL for Spanish captioning and retrieval-augmented Gemini 2.5 Flash for target-language translation, achieving significant improvements over the baseline.
This paper introduces a direct sign-to-sign translation model that bypasses intermediate text by using back-translation to create synthetic parallel sign language data, achieving significant improvements in speed and accuracy over cascade methods for ASL, CSL, and DGS.
This paper introduces Agentic AI Translate, an agentic translator prototype that replaces the traditional text-in/text-out paradigm with a four-stage agentic cycle grounded in translation studies theory, aiming to reframe translation as communication design rather than text conversion.
This paper applies Group Relative Policy Optimization (GRPO) to encoder-decoder Seq2Seq models for machine translation fine-tuning, using reference-free rewards (LaBSE and COMET-Kiwi) that require no parallel data, and achieves consistent improvements across 13 languages.
This paper introduces ForMaT, a parallel corpus of 3,956 PDFs across 15 language pairs designed for visually-grounded multilingual translation, preserving layout metadata to benchmark layout-aware MT systems.
This paper demonstrates that small open-weight LLMs (<30B parameters) can achieve competitive interpretable translation quality estimation, including MQM error annotations and corrections, rivaling much larger proprietary models while preserving data privacy.
This paper empirically examines the tradeoff between fluency and faithfulness in literary translation using 130,486 paragraphs from 106 novels, finding a consistent negative correlation for human and Google Translate translations, but weaker for TranslateGemma.
Compares DeepL, Gemini with basic prompt, and Gemini with glossary-augmented prompting for translating rock art Spanish-English terminology, finding that glossary-augmented prompting achieves the highest terminology accuracy (81.4%).
This paper proposes using reinforcement learning with semantic rewards (via GRPO) to expand LLMs to low-resource languages without the typical alignment tax of catastrophic forgetting, showing improved semantic quality and transferability over supervised fine-tuning.
This paper presents a case study on visually-guided movie subtitle translation for low-resource Indic languages, demonstrating that selective visual grounding improves translation quality while addressing temporal misalignment challenges.
A human review of TranslateGemma-12b's translations revealed that 71% of segments rated clean by automated metrics actually contained errors, highlighting significant gaps in metric-only evaluation for multilingual translation quality.
Tencent's AngelSlim team released Hy-MT1.5-1.8B-1.25bit, a highly compressed 1.25-bit machine translation model supporting 33 languages that fits in 440MB for on-device use. It utilizes the Sherry quantization algorithm to achieve world-class translation quality comparable to much larger models.
Researchers develop KokborokMT, a neural MT system for the low-resource Kokborok language, achieving BLEU scores of 17.30 en→trp and 38.56 trp→en by fine-tuning NLLB-200 on a 36k-sentence parallel corpus.
ReflectMT introduces a two-stage RL method that trains LRMs to internalize reflection, enabling single-pass high-quality translation with 94% fewer tokens than multi-step reasoning models like DeepSeek-R1.
Researchers from Tianjin University and Alibaba Group propose EA-RLVR, a reinforcement learning framework with verifiable rewards that improves cross-cultural entity translation in LLMs by activating parametric knowledge already encoded during pre-training, without relying on external knowledge bases. Training on 7k samples boosts Qwen3-14B's entity translation accuracy from 23.66% to 31.87% on unseen entities.
CLewR introduces a curriculum learning strategy with restarts for improving machine translation performance in LLMs through preference optimization. The method addresses catastrophic forgetting by iterating easy-to-hard curriculum multiple times, showing consistent gains across Gemma2, Qwen2.5, and Llama3.1 models.
LibreTranslate is a free and open source machine translation API that can be self-hosted, using Argos Translate instead of proprietary services.