Cross-Prompt Encoder for Low-Performing Languages. Mikaberidze, B., Saghinadze, T., Ostermann, S., & Müller, P. In Inui, K., Sakti, S., Wang, H., Wong, D. F., Bhattacharyya, P., Banerjee, B., Ekbal, A., Chakraborty, T., & Singh, D. P., editors, Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics, pages 2380–2393, Mumbai, India, December, 2025. The Asian Federation of Natural Language Processing and The Association for Computational Linguistics. ![Cross-Prompt Encoder for Low-Performing Languages [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper abstract bibtex Soft prompts have emerged as a powerful alternative to adapters in parameter-efficient fine-tuning (PEFT), enabling large language models (LLMs) to adapt to downstream tasks without architectural changes or parameter updates. While prior work has focused on stabilizing training via parameter interaction in small neural prompt encoders, their broader potential for transfer across languages remains unexplored. In this paper, we demonstrate that a prompt encoder can play a central role in improving performance on low-performing languages—those that achieve poor accuracy even under full-model fine-tuning. We investigate a lightweight encoder paired with multi-source training on typologically diverse languages. We call this architecture-training combination the Cross-Prompt Encoder (XPE), and show that it advances the capture of abstract, transferable patterns across languages. To complement XPE, we propose a Dual Soft Prompt mechanism that combines an encoder-based prompt with a directly trained standard soft prompt. This hybrid design proves especially effective for target languages that benefit from both broadly shared structure and language-specific alignment. Text classification experiments with a transformer encoder (XLM-R) on the SIB-200 benchmark reveal a consistent trade-off: XPE is most effective for low-performing languages, while hybrid variants offer broader adaptability across multilingual settings.
@inproceedings{mikaberidzeCrossPromptEncoderLowPerforming2025,
address = {Mumbai, India},
title = {Cross-{Prompt} {Encoder} for {Low}-{Performing} {Languages}},
isbn = {979-8-89176-303-6},
url = {https://aclanthology.org/2025.findings-ijcnlp.144/},
abstract = {Soft prompts have emerged as a powerful alternative to adapters in parameter-efficient fine-tuning (PEFT), enabling large language models (LLMs) to adapt to downstream tasks without architectural changes or parameter updates. While prior work has focused on stabilizing training via parameter interaction in small neural prompt encoders, their broader potential for transfer across languages remains unexplored. In this paper, we demonstrate that a prompt encoder can play a central role in improving performance on low-performing languages—those that achieve poor accuracy even under full-model fine-tuning. We investigate a lightweight encoder paired with multi-source training on typologically diverse languages. We call this architecture-training combination the Cross-Prompt Encoder (XPE), and show that it advances the capture of abstract, transferable patterns across languages. To complement XPE, we propose a Dual Soft Prompt mechanism that combines an encoder-based prompt with a directly trained standard soft prompt. This hybrid design proves especially effective for target languages that benefit from both broadly shared structure and language-specific alignment. Text classification experiments with a transformer encoder (XLM-R) on the SIB-200 benchmark reveal a consistent trade-off: XPE is most effective for low-performing languages, while hybrid variants offer broader adaptability across multilingual settings.},
urldate = {2026-02-26},
booktitle = {Proceedings of the 14th {International} {Joint} {Conference} on {Natural} {Language} {Processing} and the 4th {Conference} of the {Asia}-{Pacific} {Chapter} of the {Association} for {Computational} {Linguistics}},
publisher = {The Asian Federation of Natural Language Processing and The Association for Computational Linguistics},
author = {Mikaberidze, Beso and Saghinadze, Temo and Ostermann, Simon and Müller, Philipp},
editor = {Inui, Kentaro and Sakti, Sakriani and Wang, Haofen and Wong, Derek F. and Bhattacharyya, Pushpak and Banerjee, Biplab and Ekbal, Asif and Chakraborty, Tanmoy and Singh, Dhirendra Pratap},
month = dec,
year = {2025},
pages = {2380--2393},
}
Downloads: 0
{"_id":"gZwS8QwLgTfpkKAfv","bibbaseid":"mikaberidze-saghinadze-ostermann-mller-crosspromptencoderforlowperforminglanguages-2025","author_short":["Mikaberidze, B.","Saghinadze, T.","Ostermann, S.","Müller, P."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","address":"Mumbai, India","title":"Cross-Prompt Encoder for Low-Performing Languages","isbn":"979-8-89176-303-6","url":"https://aclanthology.org/2025.findings-ijcnlp.144/","abstract":"Soft prompts have emerged as a powerful alternative to adapters in parameter-efficient fine-tuning (PEFT), enabling large language models (LLMs) to adapt to downstream tasks without architectural changes or parameter updates. While prior work has focused on stabilizing training via parameter interaction in small neural prompt encoders, their broader potential for transfer across languages remains unexplored. In this paper, we demonstrate that a prompt encoder can play a central role in improving performance on low-performing languages—those that achieve poor accuracy even under full-model fine-tuning. We investigate a lightweight encoder paired with multi-source training on typologically diverse languages. We call this architecture-training combination the Cross-Prompt Encoder (XPE), and show that it advances the capture of abstract, transferable patterns across languages. To complement XPE, we propose a Dual Soft Prompt mechanism that combines an encoder-based prompt with a directly trained standard soft prompt. This hybrid design proves especially effective for target languages that benefit from both broadly shared structure and language-specific alignment. Text classification experiments with a transformer encoder (XLM-R) on the SIB-200 benchmark reveal a consistent trade-off: XPE is most effective for low-performing languages, while hybrid variants offer broader adaptability across multilingual settings.","urldate":"2026-02-26","booktitle":"Proceedings of the 14th International Joint Conference on Natural Language Processing and the 4th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics","publisher":"The Asian Federation of Natural Language Processing and The Association for Computational Linguistics","author":[{"propositions":[],"lastnames":["Mikaberidze"],"firstnames":["Beso"],"suffixes":[]},{"propositions":[],"lastnames":["Saghinadze"],"firstnames":["Temo"],"suffixes":[]},{"propositions":[],"lastnames":["Ostermann"],"firstnames":["Simon"],"suffixes":[]},{"propositions":[],"lastnames":["Müller"],"firstnames":["Philipp"],"suffixes":[]}],"editor":[{"propositions":[],"lastnames":["Inui"],"firstnames":["Kentaro"],"suffixes":[]},{"propositions":[],"lastnames":["Sakti"],"firstnames":["Sakriani"],"suffixes":[]},{"propositions":[],"lastnames":["Wang"],"firstnames":["Haofen"],"suffixes":[]},{"propositions":[],"lastnames":["Wong"],"firstnames":["Derek","F."],"suffixes":[]},{"propositions":[],"lastnames":["Bhattacharyya"],"firstnames":["Pushpak"],"suffixes":[]},{"propositions":[],"lastnames":["Banerjee"],"firstnames":["Biplab"],"suffixes":[]},{"propositions":[],"lastnames":["Ekbal"],"firstnames":["Asif"],"suffixes":[]},{"propositions":[],"lastnames":["Chakraborty"],"firstnames":["Tanmoy"],"suffixes":[]},{"propositions":[],"lastnames":["Singh"],"firstnames":["Dhirendra","Pratap"],"suffixes":[]}],"month":"December","year":"2025","pages":"2380–2393","bibtex":"@inproceedings{mikaberidzeCrossPromptEncoderLowPerforming2025,\n\taddress = {Mumbai, India},\n\ttitle = {Cross-{Prompt} {Encoder} for {Low}-{Performing} {Languages}},\n\tisbn = {979-8-89176-303-6},\n\turl = {https://aclanthology.org/2025.findings-ijcnlp.144/},\n\tabstract = {Soft prompts have emerged as a powerful alternative to adapters in parameter-efficient fine-tuning (PEFT), enabling large language models (LLMs) to adapt to downstream tasks without architectural changes or parameter updates. While prior work has focused on stabilizing training via parameter interaction in small neural prompt encoders, their broader potential for transfer across languages remains unexplored. In this paper, we demonstrate that a prompt encoder can play a central role in improving performance on low-performing languages—those that achieve poor accuracy even under full-model fine-tuning. We investigate a lightweight encoder paired with multi-source training on typologically diverse languages. We call this architecture-training combination the Cross-Prompt Encoder (XPE), and show that it advances the capture of abstract, transferable patterns across languages. To complement XPE, we propose a Dual Soft Prompt mechanism that combines an encoder-based prompt with a directly trained standard soft prompt. This hybrid design proves especially effective for target languages that benefit from both broadly shared structure and language-specific alignment. Text classification experiments with a transformer encoder (XLM-R) on the SIB-200 benchmark reveal a consistent trade-off: XPE is most effective for low-performing languages, while hybrid variants offer broader adaptability across multilingual settings.},\n\turldate = {2026-02-26},\n\tbooktitle = {Proceedings of the 14th {International} {Joint} {Conference} on {Natural} {Language} {Processing} and the 4th {Conference} of the {Asia}-{Pacific} {Chapter} of the {Association} for {Computational} {Linguistics}},\n\tpublisher = {The Asian Federation of Natural Language Processing and The Association for Computational Linguistics},\n\tauthor = {Mikaberidze, Beso and Saghinadze, Temo and Ostermann, Simon and Müller, Philipp},\n\teditor = {Inui, Kentaro and Sakti, Sakriani and Wang, Haofen and Wong, Derek F. and Bhattacharyya, Pushpak and Banerjee, Biplab and Ekbal, Asif and Chakraborty, Tanmoy and Singh, Dhirendra Pratap},\n\tmonth = dec,\n\tyear = {2025},\n\tpages = {2380--2393},\n}\n\n\n\n","author_short":["Mikaberidze, B.","Saghinadze, T.","Ostermann, S.","Müller, P."],"editor_short":["Inui, K.","Sakti, S.","Wang, H.","Wong, D. F.","Bhattacharyya, P.","Banerjee, B.","Ekbal, A.","Chakraborty, T.","Singh, D. P."],"key":"mikaberidzeCrossPromptEncoderLowPerforming2025","id":"mikaberidzeCrossPromptEncoderLowPerforming2025","bibbaseid":"mikaberidze-saghinadze-ostermann-mller-crosspromptencoderforlowperforminglanguages-2025","role":"author","urls":{"Paper":"https://aclanthology.org/2025.findings-ijcnlp.144/"},"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://bibbase.org/zotero-group/ndrmyrvtl/5786256","dataSources":["5EQ7rDLpRYyekvrfx"],"keywords":[],"search_terms":["cross","prompt","encoder","low","performing","languages","mikaberidze","saghinadze","ostermann","müller"],"title":"Cross-Prompt Encoder for Low-Performing Languages","year":2025}