List your company

Machine Translation (EN->DE) on IWSLT 2017 (test)

37.1BLEUFAAST
20.56424.85729.1533.443Jun 5, 2024Sep 29, 2024Jan 24, 2025May 21, 2025Sep 14, 2025Jan 9, 2026May 6, 2026

Evaluation results

MethodLinks
FAASTBackbone=Qwen2.5-7B-Instruct, Protocol=FAAST, Data Setting=full2026.0537.1———
FAASTBackbone=Qwen2.5-3B-Instruct, Protocol=FAAST, Data Setting=full2026.0536.4———
oursStudent Model Architecture=T5-Large, Teacher Model=T5-XL2024.0636———
In-Context LearningBackbone=Qwen2.5-7B-Instruct, Protocol=In-Context Learning, Data Setting=1-shot2026.0535.7———
DistiLLMStudent Model Architecture=T5-Large, Teacher Model=T5-XL2024.0635.5———
FAASTBackbone=Qwen2.5-7B-Instruct, Protocol=FAAST, Data Setting=1-shot2026.0535.34———
T5-XL (teacher)Model Architecture=T5-XL2024.0635.2———
GKDStudent Model Architecture=T5-Large, Teacher Model=T5-XL2024.0634.7———
Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct, Protocol=zero-shot, Data Setting=full2026.0534.69———
ImitKDStudent Model Architecture=T5-Large, Teacher Model=T5-XL2024.0634.1———
SFT (student)Student Model Architecture=T5-Large, Teacher Model=T5-XL2024.0633.7———
FAASTBackbone=Qwen2.5-3B-Instruct, Protocol=FAAST, Data Setting=1-shot2026.0533.23———
Qwen2.5-3B-InstructBackbone=Qwen2.5-3B-Instruct, Protocol=zero-shot, Data Setting=full2026.0532.92———
SeqKDStudent Model Architecture=T5-Large, Teacher Model=T5-XL2024.0632.9———
oursStudent Model Architecture=T5-Base, Teacher Model=T5-XL2024.0632.4———
Google translatorTrained Model=-2025.0532.3980.29268.2
In-Context LearningBackbone=Qwen2.5-3B-Instruct, Protocol=In-Context Learning, Data Setting=1-shot2026.0532.33———
GPT4.1-nanoTrained Model=-2025.0531.8983.99670.6
KDStudent Model Architecture=T5-Large, Teacher Model=T5-XL2024.0631.7———
GKDStudent Model Architecture=T5-Base, Teacher Model=T5-XL2024.0631.2———
DistiLLMStudent Model Architecture=T5-Base, Teacher Model=T5-XL2024.0630.8———
SFT (student)Student Model Architecture=T5-Base, Teacher Model=T5-XL2024.0630.1———
KDStudent Model Architecture=T5-Base, Teacher Model=T5-XL2024.0629.8———
SeqKDStudent Model Architecture=T5-Base, Teacher Model=T5-XL2024.0629.2———
ImitKDStudent Model Architecture=T5-Base, Teacher Model=T5-XL2024.0628.7———
FAASTBackbone=Qwen2.5-7B-Instruct, Protocol=FAAST, Data Setting=full2026.0527.75———
UniRBackbone=Llama3.2-3B, Trained Model=1B2025.0527.0381.4393.767.39
FAASTBackbone=Qwen2.5-7B-Instruct, Protocol=FAAST, Data Setting=1-shot2026.0526.77———
Qwen2.5-7B-InstructBackbone=Qwen2.5-7B-Instruct, Protocol=zero-shot, Data Setting=full2026.0525.53———
In-Context LearningBackbone=Qwen2.5-7B-Instruct, Protocol=In-Context Learning, Data Setting=1-shot2026.0525.39———
GRPO FullBackbone=Llama3.2-3B, Trained Model=3B2025.0525.2680.1392.8866.09
FAASTBackbone=Qwen2.5-3B-Instruct, Protocol=FAAST, Data Setting=full2026.0525.22———
GRPO LoRABackbone=Llama3.2-3B, Trained Model=3B2025.0524.1878.9391.9765.03
Backbone + 1BBackbone=Llama3.2-3B, Trained Model=-2025.0524.178.791.4764.76
oursStudent Model Architecture=T5-Small, Teacher Model=T5-XL2024.0623.7———
FAASTBackbone=Qwen2.5-3B-Instruct, Protocol=FAAST, Data Setting=1-shot2026.0523.35———
Qwen2.5-3B-InstructTrained Model=-2025.0523.2878.2591.2664.26
Qwen2.5-3B-InstructBackbone=Qwen2.5-3B-Instruct, Protocol=zero-shot, Data Setting=full2026.0523.22———
Backbone onlyBackbone=Llama3.2-3B, Trained Model=-2025.0523.1977.7290.9763.96
In-Context LearningBackbone=Qwen2.5-3B-Instruct, Protocol=In-Context Learning, Data Setting=1-shot2026.0523.03———
GKDStudent Model Architecture=T5-Small, Teacher Model=T5-XL2024.0622.7———
DistiLLMStudent Model Architecture=T5-Small, Teacher Model=T5-XL2024.0622.5———
ImitKDStudent Model Architecture=T5-Small, Teacher Model=T5-XL2024.0622.2———
KDStudent Model Architecture=T5-Small, Teacher Model=T5-XL2024.0621.7———
SFT (student)Student Model Architecture=T5-Small, Teacher Model=T5-XL2024.0621.5———
SeqKDStudent Model Architecture=T5-Small, Teacher Model=T5-XL2024.0621.2———