Modelpublic

MDK-YLC/Qwen3.5-2B-WMT26-Ukrainian-LoRA

sourceHugging Facemitupdated 1mo agoView on Hugging Face

0likes8downloads

Model Card

Zolint at WMT 2026: A Multitask LLM Submission for the Low-Resource Ukrainian Track


Team	Zolint
Institution	UTokyo & TUM
This adapter	`MDK-YLC/Qwen3.5-2B-WMT26-Ukrainian-LoRA`
Base model	`Qwen/Qwen3.5-2B`
Code	`vanessafu/nlp-2026-low-resource-ukrainian-mt-qa`

LoRA adapter for the Ukrainian track of the WMT 2026 Shared Task on Multitask LLMs with Limited Resources (MT, QA, SC, GC, MR).

On the official test set, our unified system obtains the best score on all reported metrics except question answering.

Official test highlights

System	MT All chrF++	CS-UK BLEU	EN-UK BLEU	QA All EM	SC All EM	GC All EM	Math EM
Baseline	18.46	5.38	0.55	36.33	8.38	2.18	20.00
TUMHN	16.08	3.93	1.01	40.13	16.50	6.80	4.40
Koshi	16.80	6.01	0.70	46.99	52.50	30.55	0.40
Ours (Zolint)	49.86	23.66	15.43	41.15	66.43	45.18	21.60

Training summary

1.Stage 1: LoRA (r=16, α=32) on MT (en→uk) + GC + SC; keep best epoch by combined MT+GC+SC.
2.Stage 2: Continue 1 epoch on QA + MR + MT/GC replay + full Czech→Ukrainian chat data.

Quick start

bash

pip install torch transformers peft accelerate

python

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_id = "Qwen/Qwen3.5-2B"
lora_id = "MDK-YLC/Qwen3.5-2B-WMT26-Ukrainian-LoRA"

tokenizer = AutoTokenizer.from_pretrained(lora_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    base_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True,
)
model = PeftModel.from_pretrained(model, lora_id)
model.eval()

See inference_example.py in this repo for MT / QA / SC-style prompts.

Files

File	Description
`adapter_config.json`	LoRA / PEFT config
`adapter_model.safetensors`	LoRA weights
`tokenizer.json` / `tokenizer_config.json`	Tokenizer
`chat_template.jinja`	Chat template
`inference_example.py`	Minimal runnable example

Note: Adapter only — base weights come from Qwen/Qwen3.5-2B.

Citation

Zolint at WMT 2026: A Multitask LLM Submission for the Low-Resource Ukrainian Track Hugging Face: MDK-YLC/Qwen3.5-2B-WMT26-Ukrainian-LoRA