This project file contains:

1. Requirements file: -requirements.txt **install requirements by pip install -r requirements.txt **

2. Three baseline models: ## The three baseline model are from PAN challenge official document and were modified a bit run them by: run xxx.py -backtranslation_baseline_clef25.py -baseline_delete_clef25.py -mt0_baseline_clef25.py

3. Qwen3 based models with data preparation codes: ## The four Qwen3 based models are for our project -qwen3_14b_0shot.py zero shot experiment with Qwen3-14B run by command: run qwen3_14b_0shot.py

   -qwen3_14b_fewshot.py few shot experiment with Qwen3-14B, randomly select 3 examples for each language as shots run by command: run qwen3_14b_fewshot.py

   -qwen3_14b_rag.py -build_multi_index.py train Qwen3-14B with RAG strategy before running Qwen3-14B model, prepare the embedding of traing data via: run build_multi_index.py run RAG model by command: run qwen3_14b_rag.py

   -qwen3_14b_finetune_LoRA_peft.py -finetune_prepare.py finetune Qwen3-14B with Low Rank Adaptation and peft strategy prepare the training data via: run finetune_prepare.py run finetuning model by command: run qwen3_14b_finetune_LpRA_peft.py

4. Dataset: ## the dataset is a document, de refers to German, en refers to English, es refers to Spanish and zh refers to Chinese ## xx_data.tsv: full dataset ## xx_train.tsv: training set(70%) ## xx_dev.tsv: development set(15%) ## xx_test.tsv: test set(15%) =input_data -de_data.tsv -de_train.tsv -de_dev.tsv -de_test.tsv -en_data.tsv -en_train.tsv -en_dev.tsv -en_test.tsv -es_data.tsv -es_train.tsv -es_dev.tsv -es_test.tsv -zh_data.tsv -zh_train.tsv -zh_dev.tsv -zh_test.tsv

5. Evaluations: ## evaluation method is adopted from PAN challenges ==evaluation =metrics -similarity.py -toxity.py =fluency -deberta_encoder.py -xcomet.py -evaluate.py -utils.py run the evaluation by command: run evaluate.py