Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Tower+: Bridging Generality and Translation Specialization in Multilingual LLMs

About

Fine-tuning pretrained LLMs has been shown to be an effective strategy for reaching state-of-the-art performance on specific tasks like machine translation. However, this process of adaptation often implies sacrificing general-purpose capabilities, such as conversational reasoning and instruction-following, hampering the utility of the system in real-world applications that require a mixture of skills. In this paper, we introduce Tower+, a suite of models designed to deliver strong performance across both translation and multilingual general-purpose text capabilities. We achieve a Pareto frontier between translation specialization and multilingual general-purpose capabilities by introducing a novel training recipe that builds on Tower (Alves et al., 2024), comprising continued pretraining, supervised fine-tuning, preference optimization, and reinforcement learning with verifiable rewards. At each stage of training, we carefully generate and curate data to strengthen performance on translation as well as general-purpose tasks involving code generation, mathematics problem solving, and general instruction-following. We develop models at multiple scales: 2B, 9B, and 72B. Our smaller models often outperform larger general-purpose open-weight and proprietary LLMs (e.g., Llama 3.3 70B, GPT-4o). Our largest model delivers best-in-class translation performance for high-resource languages and top results in multilingual Arena Hard evaluations and in IF-MT, a benchmark we introduce for evaluating both translation and instruction-following. Our findings highlight that it is possible to rival frontier models in general capabilities, while optimizing for specific business domains, such as translation and localization.

Ricardo Rei, Nuno M. Guerreiro, Jos\'e Pombal, Jo\~ao Alves, Pedro Teixeirinha, Amin Farajian, Andr\'e F. T. Martins• 2025

Related benchmarks

TaskDatasetResultRank
Machine TranslationFLORES+ (test)
spBLEU45.32
128
Machine TranslationWMT24++ v1.0 (test)
XCOMET Score88.19
49
Machine Translation (xx -> zh)FLORES+ latest (test)
spBLEU33.25
30
Machine TranslationWMT 2025 (test)
XCOMET-XXL41
17
Machine TranslationFLORES-200 EN ⇔ XX 2022
XCOMET-XXL84.16
17
Machine TranslationFLORES-200 ZH ⇔ XX 2022
XCOMET-XXL0.7969
17
Machine TranslationFLORES-200 XX ⇔ XX 2022
XCOMET-XXL70.02
17
Machine TranslationMandarin ⇔ Minority (test)
XCOMET-XXL0.3855
16
Showing 8 of 8 rows

Other info

Follow for update