Browse State-of-the-Art › Machine Translation

Machine Translation

2,444 papers with code · 84 benchmarks · 82 datasets archive 2025-07-28

Natural Language Processing

Machine translation is the task of translating a sentence in a source language to a different target language.

Approaches for machine translation can range from rule-based to statistical to neural-based. More recently, encoder-decoder attention-based architectures like BERT have attained major improvements in machine translation.

One of the most popular datasets used to benchmark machine translation systems is the WMT family of datasets. Some of the most commonly used evaluation metrics for machine translation systems include BLEU, METEOR, NIST, and others.

( Image credit: Google seq2seq )

Description from the archive archive 2025-07-28.

Benchmarks archive 2025-07-28

84 leaderboard tables shown for this task, 84 with rows (a “benchmark” on this site is a table with at least one row, as on /sota), ordered by row count. “Best model” is the first row in the archive's own order at snapshot; nothing is re-ranked here and metric direction is not recorded in the archive. PwC's Trend sparklines are not in the archive, so that column is omitted. 10 shown of 84 until expanded.

Syntology 3 newer papers were judged to report on this task's benchmarks so far (Claude Sonnet 4.5, reading their tables: a model's judgement, not a result; 0 placed): WMT2014 German-English (1 judged, 0 placed), FLoRes-200 (2 judged, 0 placed). Drawn from 8,886 of 18,637 newer papers checked so far (733 more had nothing to place; 41 could not be read).

DatasetBest model (first row in archive order)PaperCodeSyntologyCompare
WMT2014 English-German (91 rows) Transformer Cycle (Rev) Lessons on Parameter Sharing across Layers in Transformers code Syntology official (archive's flag): 2 ran · 2 ran (of which 2 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified; every one of the 2 samples that ran constructed an object rather than computing a result Compare
WMT2014 English-French (57 rows) Transformer+BT (ADMIN init) Very Deep Transformers for Neural Machine Translation code Syntology community repositories only · 8 ran (of which 2 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 5 where Syntology's instrument failed) · 1 unverified Compare
IWSLT2014 German-English (34 rows) PiNMT Integrating Pre-trained Language Model into Neural Machine Translation — — Compare
ACES (21 rows) HWTSC-Teacher-Sim ACES: Translation Accuracy Challenge Sets for Evaluating Machine... code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 1 honoured, 0 violated, 0 with no contract checked; 1 where Syntology's instrument failed) · 0 unverified Compare
WMT2016 English-Romanian (21 rows) DeLighT DeLighT: Deep and Light-weight Transformer code Syntology official (archive's flag): 3 ran · 3 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified Compare
WMT2016 Romanian-English (21 rows) fast-noisy-channel-modeling Language Models not just for Pre-training: Fast Online Neural... code — Compare
WMT2014 German-English (16 rows) Bi-SimCut Bi-SimCut: A Simple Strategy for Boosting Neural Machine Translation code — Compare
IWSLT2015 German-English (15 rows) PS-KD Self-Knowledge Distillation with Progressive Refinement of Targets code Syntology official (archive's flag): 5 ran · 5 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 4 where Syntology's instrument failed) · 0 unverified Compare
WMT2016 English-German (12 rows) MADL Multi-Agent Dual Learning — — Compare
IWSLT2015 English-Vietnamese (11 rows) EnViT5 + MTet MTet: Multi-domain Translation for English and Vietnamese code — Compare
IWSLT2015 English-German (8 rows) PS-KD Self-Knowledge Distillation with Progressive Refinement of Targets code Syntology official (archive's flag): 5 ran · 5 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 4 where Syntology's instrument failed) · 0 unverified Compare
WMT2016 German-English (8 rows) FLAN 137B (few-shot, k=11) Finetuned Language Models Are Zero-Shot Learners code Syntology official: harvested for another paper · 0 ran · 1 unverified Compare
IWSLT2014 English-German (6 rows) PiNMT Integrating Pre-trained Language Model into Neural Machine Translation — — Compare
WMT2015 English-German (6 rows) ByteNet Neural Machine Translation in Linear Time code — Compare
FLoRes-200 (5 rows) GenTranslate-7B GenTranslate: Large Language Models are Generative Multilingual... code — Compare
WMT 2017 Latvian-English (4 rows) Transformer trained on highly filtered data Impact of Corpora Quality on Neural Machine Translation code — Compare
WMT2016 English-Russian (4 rows) Attentional encoder-decoder + BPE Edinburgh Neural Machine Translation Systems for WMT 16 code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified Compare
flores95-devtest eng-X (3 rows) SeamlessM4T Large SeamlessM4T: Massively Multilingual & Multimodal Machine Translation code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified Compare
flores95-devtest X-eng (3 rows) SeamlessM4T Large SeamlessM4T: Massively Multilingual & Multimodal Machine Translation code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified Compare
FRMT (Chinese - Mainland) (3 rows) PaLM 2 PaLM 2 Technical Report code — Compare
FRMT (Chinese - Taiwan) (3 rows) PaLM 2 PaLM 2 Technical Report code — Compare
FRMT (Portuguese - Portugal) (3 rows) PaLM 2 PaLM 2 Technical Report code — Compare
FRMT (Portuguese - Brazil) (3 rows) PaLM 2 PaLM 2 Technical Report code — Compare
WMT 2017 English-Chinese (3 rows) DynamicConv Pay Less Attention with Lightweight and Dynamic Convolutions code — Compare
WMT2014 French-English (3 rows) FLAN 137B (few-shot, k=9) Finetuned Language Models Are Zero-Shot Learners code Syntology official: harvested for another paper · 0 ran · 1 unverified Compare
WMT2017 Chinese-English (3 rows) StrokeNet Breaking the Representation Bottleneck of Chinese Characters:... code — Compare
20NEWS (2 rows) tensorflow/tensor2tensor Neural Machine Translation code — Compare
Arba Sicula (2 rows) Larger Sicilian Translator: A Recipe for Low-Resource NMT code — Compare
Itihasa (2 rows) Baseline (en->sn) Itihasa: A large-scale corpus for Sanskrit to English translation — — Compare
IWSLT2015 Vietnamese-English (2 rows) HeadMask (Random-18) Alleviating the Inequality of Attention Heads for Neural Machine... — — Compare
IWSLT2017 Arabic-English (2 rows) Transformer base + BPE-Dropout BPE-Dropout: Simple and Effective Subword Regularization code Syntology official (archive's flag): 3 ran · 24 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 0 honoured, 1 violated, 14 with no contract checked; 9 where Syntology's instrument failed) · 0 unverified Compare
IWSLT2017 English-French (2 rows) Transformer base + BPE-Dropout BPE-Dropout: Simple and Effective Subword Regularization code Syntology official (archive's flag): 3 ran · 24 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 0 honoured, 1 violated, 14 with no contract checked; 9 where Syntology's instrument failed) · 0 unverified Compare
IWSLT2017 English-Arabic (2 rows) Transformer base + BPE-Dropout BPE-Dropout: Simple and Effective Subword Regularization code Syntology official (archive's flag): 3 ran · 24 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 0 honoured, 1 violated, 14 with no contract checked; 9 where Syntology's instrument failed) · 0 unverified Compare
IWSLT2017 French-English (2 rows) Transformer base + BPE-Dropout BPE-Dropout: Simple and Effective Subword Regularization code Syntology official (archive's flag): 3 ran · 24 ran (of which 0 constructed an object rather than computing a result; 15 with no instrument failure: 0 honoured, 1 violated, 14 with no contract checked; 9 where Syntology's instrument failed) · 0 unverified Compare
IWSLT2017 German-English (2 rows) Adaptively Sparse Transformer (alpha-entmax) Adaptively Sparse Transformers code Syntology official (archive's flag): 2 ran · 3 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 3 where Syntology's instrument failed) · 0 unverified Compare
WMT 2018 Finnish-English (2 rows) CT+B/S construction The University of Sydney's Machine Translation System for WMT19 — — Compare
WMT2014 English-Czech (2 rows) Evolved Transformer Big The Evolved Transformer code — Compare
WMT2017 Turkish-English (2 rows) HeadMask (Random-18) Alleviating the Inequality of Attention Heads for Neural Machine... — — Compare
WMT2019 English-German (2 rows) Facebook FAIR (ensemble) Facebook FAIR's WMT19 News Translation Task Submission code Syntology 7 ran (of which 0 constructed an object rather than computing a result; 6 with no instrument failure: 0 honoured, 0 violated, 6 with no contract checked; 1 where Syntology's instrument failed) · 2 unverified Compare
ACCURAT balanced test corpus for under resourced languages Russian-Estonian (1 row) Multilingual Transformer Training and Adapting Multilingual NMT for Less-resourced and... code — Compare
ACCURAT balanced test corpus for under resourced languages Estonian-Russian (1 row) Multilingual Transformer Training and Adapting Multilingual NMT for Less-resourced and... code — Compare
Alexa Point of View (1 row) T5 Converting the Point of View of Messages Spoken to Virtual Assistants code — Compare
Business Scene Dialogue JA-EN (1 row) Transformer-base Designing the Business Conversation Corpus code — Compare
Business Scene Dialogue EN-JA (1 row) Transformer-base Designing the Business Conversation Corpus code — Compare
IWSLT 2017 (1 row) GPT-4o (HPT) Hierarchical Prompting Taxonomy: A Universal Evaluation Framework... code — Compare
IWSLT2015 Chinese-English (1 row) BP-Transformer BP-Transformer: Modelling Long-Range Context via Binary Partitioning code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 2 with no instrument failure: 0 honoured, 0 violated, 2 with no contract checked; 0 where Syntology's instrument failed) · 5 unverified Compare
IWSLT2015 Thai-English (1 row) Seq-KD + Seq-Inter + Word-KD Sequence-Level Knowledge Distillation code Syntology official (archive's flag): 1 ran · 1 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 0 where Syntology's instrument failed) · 1 unverified Compare
Multi Lingual Bug Reports (1 row) ChatGPT English Please: Evaluating Machine Translation with Large Language... code — Compare
OpenSubtitles (1 row) Fine tuned MarianMT Crossing Language Borders: A Pipeline for Indonesian Manhwa Translation code — Compare
slone/myv_ru_2022 myv-ru (1 row) slone/mbart-large-51-myv-mul-v1 The first neural machine translation system for the Erzya language code — Compare
slone/myv_ru_2022 ru-myv (1 row) slone/mbart-large-51-mul-myv-v1 The first neural machine translation system for the Erzya language code — Compare
Tatoeba (EL-to-EN) (1 row) PENELOPIE (Transformers-based Greek-to-English NMT) PENELOPIE: Enabling Open Information Extraction for the Greek... code — Compare
Tatoeba (EN-to-EL) (1 row) PENELOPIE Transformers-based NMT (EN2EL) PENELOPIE: Enabling Open Information Extraction for the Greek... code — Compare
V_A (trained on T_H) (1 row) M_C On Automatic Parsing of Log Records code — Compare
V_B (trained on T_H) (1 row) M_C On Automatic Parsing of Log Records code — Compare
V_C (trained on T_H) (1 row) M_C On Automatic Parsing of Log Records code — Compare
WMT 2017 English-Latvian (1 row) Transformer trained on highly filtered data Impact of Corpora Quality on Neural Machine Translation code — Compare
WMT 2018 English-Estonian (1 row) Multi-pass backtranslated adapted transformer Tilde's Machine Translation Systems for WMT 2018 code — Compare
WMT 2018 English-Finnish (1 row) Transformer trained on highly filtered data Impact of Corpora Quality on Neural Machine Translation code — Compare
WMT 2018 Estonian-English (1 row) Multi-pass backtranslated adapted transformer Tilde's Machine Translation Systems for WMT 2018 code — Compare
WMT 2022 Chinese-English (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 Czech-English (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 English-German (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 English-Japanese (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 English-Russian (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 English-Chinese (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 English-Czech (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 German-English (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 Japanese-English (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT 2022 Russian-English (1 row) Vega-MT Vega-MT: The JD Explore Academy Translation System for WMT22 code — Compare
WMT2015 English-Russian (1 row) C2-50k Segmentation Neural Machine Translation of Rare Words with Subword Units code Syntology official (archive's flag): 4 ran · 21 ran (of which 0 constructed an object rather than computing a result; 1 with no instrument failure: 0 honoured, 0 violated, 1 with no contract checked; 20 where Syntology's instrument failed) · 9 unverified Compare
WMT2016 Czech-English (1 row) Attentional encoder-decoder + BPE Edinburgh Neural Machine Translation Systems for WMT 16 code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified Compare
WMT2016 English-Czech (1 row) Attentional encoder-decoder + BPE Edinburgh Neural Machine Translation Systems for WMT 16 code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified Compare
WMT2016 English-French (1 row) DeLighT DeLighT: Deep and Light-weight Transformer code Syntology official (archive's flag): 3 ran · 3 ran (of which 0 constructed an object rather than computing a result; 3 with no instrument failure: 0 honoured, 0 violated, 3 with no contract checked; 0 where Syntology's instrument failed) · 0 unverified Compare
WMT2016 Finnish-English (1 row) CT+B/S construction The University of Sydney's Machine Translation System for WMT19 — — Compare
WMT2016 Russian-English (1 row) Attentional encoder-decoder + BPE Edinburgh Neural Machine Translation Systems for WMT 16 code Syntology official (archive's flag): 2 ran · 2 ran (of which 0 constructed an object rather than computing a result; 0 with no instrument failure: 0 honoured, 0 violated, 0 with no contract checked; 2 where Syntology's instrument failed) · 0 unverified Compare
WMT2017 English-Finnish (1 row) OmniNetP OmniNet: Omnidirectional Representations from Transformers code — Compare
WMT2017 English-French (1 row) OmniNetP OmniNet: Omnidirectional Representations from Transformers code — Compare
WMT2017 English-German (1 row) OmniNetP OmniNet: Omnidirectional Representations from Transformers code — Compare
WMT2017 Finnish-English (1 row) CT+B/S construction The University of Sydney's Machine Translation System for WMT19 — — Compare
WMT2017 Russian-English (1 row) OmniNetP OmniNet: Omnidirectional Representations from Transformers code — Compare
WMT2019 English-Japanese (1 row) fiore Parallel Corpus Filtering via Pre-trained Language Models — — Compare
WMT2019 Finnish-English (1 row) CT+B/S construction The University of Sydney's Machine Translation System for WMT19 — — Compare
WMT2019 German-English (1 row) Exploiting Mono at Scale (single) Exploiting Monolingual Data at Scale for Neural Machine Translation — — Compare

Syntology column: samples harvested from the paper's repositories and executed on synthesized fixtures; “ran” is not a correctness claim and does not order the table. Each line reads “N ran (of which C constructed an object rather than computing a result; K with no instrument failure: H honoured, V violated, P with no contract checked; I where Syntology's instrument failed) · U unverified”; the instrument figure counts failures of Syntology's instrument, not of the code. When the archive marks a repository official for the row's paper, the line starts with that repository's state (“official” is the archive's flag, not a verdict on who wrote the code; “community repositories only” when every sample that ran came from a community repository, “official: no sample here; runs from other or unrecorded repositories” when some came from a repository the paper names or has in its text, or from none recorded); hover it for the repositories the samples that ran came from. A dash means no Syntology record for that paper, not a recorded non-run. Each “ran” line links to that paper's sample list. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. For agents: get_harvested_code_for_paper(arxiv_id) lists each paper's samples; how to connect.

Libraries

Not in the archive: the export carries no per-task library table, so there is nothing to show at snapshot 2025-07-28.

Datasets archive 2025-07-28

82 datasets whose archive record lists this task, ordered by the archive's paper count. 30 shown of 82 until expanded.

Subtasks archive 2025-07-28

10 subtasks in the archive's task tree.

Papers archive 2025-07-28

archive papers tagged: 10,752 · with a code link: 2,444 · where Syntology ran a sample: 477 (360 with a run with no instrument failure, 117 where every run was a failure of Syntology's instrument) Syntology

Show: all tagged papersonly where code ran (477 of 10,752 tagged: 360 with a run with no instrument failure, 117 where every run was a failure of Syntology's instrument)

Page 1 of 108: papers 1 to 100 of 10,752, in archive order: by repositories listed in the archive (most first), then newest first, not by stars (the archive holds no stars, so PwC's “Social” and “Latest” sorts cannot be reproduced). Papers that list no repository come after every paper that lists one.

Papers without a page here are shown as plain text. A Syntology line reads “N ran (of which C constructed an object rather than computing a result; K with no instrument failure: H honoured, V violated, P with no contract checked; I where Syntology's instrument failed) · U unverified”; the figure “where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. When the archive marks a repository official for the paper, the line starts with that repository's state (the archive's flag, not a verdict on who wrote the code); hover it for the repositories the samples that ran came from. The first sentence of the abstract is shown for the first 30 papers; every abstract is on its paper's page.

Syntology lines on 68 of the papers shown; no Syntology record for the others (a paper without an arXiv id cannot be joined to the graph, and absence from the graph layer is not a recorded non-run). “Ran” means the sample executed on a synthesized fixture, not that the paper's result was reproduced; each line links to that paper's sample list. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. For agents: get_harvested_code_for_paper(arxiv_id) lists each paper's samples; how to connect.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections