SynthDetoxM: Modern LLMs are Few-Shot Parallel Detoxification Data Annotators
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2025
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908490333683712 |
|---|---|
| author | Moskovskiy, Daniil Sushko, Nikita Pletenev, Sergey Tutubalina, Elena Panchenko, Alexander |
| author_facet | Moskovskiy, Daniil Sushko, Nikita Pletenev, Sergey Tutubalina, Elena Panchenko, Alexander |
| contents | Existing approaches to multilingual text detoxification are hampered by the scarcity of parallel multilingual datasets. In this work, we introduce a pipeline for the generation of multilingual parallel detoxification data. We also introduce SynthDetoxM, a manually collected and synthetically generated multilingual parallel text detoxification dataset comprising 16,000 high-quality detoxification sentence pairs across German, French, Spanish and Russian. The data was sourced from different toxicity evaluation datasets and then rewritten with nine modern open-source LLMs in few-shot setting. Our experiments demonstrate that models trained on the produced synthetic datasets have superior performance to those trained on the human-annotated MultiParaDetox dataset even in data limited setting. Models trained on SynthDetoxM outperform all evaluated LLMs in few-shot setting. We release our dataset and code to help further research in multilingual text detoxification. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_06394 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SynthDetoxM: Modern LLMs are Few-Shot Parallel Detoxification Data Annotators Moskovskiy, Daniil Sushko, Nikita Pletenev, Sergey Tutubalina, Elena Panchenko, Alexander Computation and Language Existing approaches to multilingual text detoxification are hampered by the scarcity of parallel multilingual datasets. In this work, we introduce a pipeline for the generation of multilingual parallel detoxification data. We also introduce SynthDetoxM, a manually collected and synthetically generated multilingual parallel text detoxification dataset comprising 16,000 high-quality detoxification sentence pairs across German, French, Spanish and Russian. The data was sourced from different toxicity evaluation datasets and then rewritten with nine modern open-source LLMs in few-shot setting. Our experiments demonstrate that models trained on the produced synthetic datasets have superior performance to those trained on the human-annotated MultiParaDetox dataset even in data limited setting. Models trained on SynthDetoxM outperform all evaluated LLMs in few-shot setting. We release our dataset and code to help further research in multilingual text detoxification. |
| title | SynthDetoxM: Modern LLMs are Few-Shot Parallel Detoxification Data Annotators |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2502.06394 |