---
title: "Dual-Axis Compositional Contrastive Few-Shot Learning using"
description: "Multilingual intent and domain classifiers usually treat each language–label pair as a separate class. 81.12% accuracy. 19 references."
authors: ["Kathakali Mitra", "Sakshi Singh", "Sree Nithish Reddy Gunapati", "Aruna Malapati", "Mark G. Lee"]
year: 2026
venue: "Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion"
source: "https://aclanthology.org/2026.ltedi-1.3/"
doi: "10.18653/v1/2026.ltedi-1.3"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/mitra-2026-dual-axis-compositional-contrastive-few/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/mitra-2026-dual-axis-compositional-contrastive-few/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/30/) / Dual-Axis Compositional Contrastive Few-Shot Learning using Prototypes Across Linguistic and Semantic Dimensions for Indic Low-Resource Multilingual NLU

# Dual-Axis Compositional Contrastive Few-Shot Learning using Prototypes Across Linguistic and Semantic Dimensions for Indic Low-Resource Multilingual NLU

**Authors:** Kathakali Mitra, Sakshi Singh, Sree Nithish Reddy Gunapati, Aruna Malapati and [Mark G. Lee](https://md.research.elamur.ai/free/en/authors/mark-lee/) · **Year:** 2026 · **Venue:** [Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion](https://md.research.elamur.ai/free/en/venues/ltedi/) ([Workshop on Language Technology for Equality, Diversity, Inclusion 2026](https://md.research.elamur.ai/free/en/venues/ltedi/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.ltedi-1.3/) · [PDF](https://aclanthology.org/2026.ltedi-1.3.pdf) · [DOI 10.18653/v1/2026.ltedi-1.3](https://doi.org/10.18653/v1/2026.ltedi-1.3) · [OpenAlex](https://openalex.org/W7166827456)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Multilingual intent and domain classifiers usually treat each language–label pair as a separate class. When a new language or a new intent arrives with only a few examples, this tends to destabilise what the model already learned. The authors address this for low-resource Indic languages by splitting the representation into two spaces: one for the language of an utterance and one for its domain–intent meaning. A mean-pooled IndicBERTv2 encoder feeds two projection heads, each with its own learnable prototype bank. Joint prototypes are built on the fly by element-wise (Hadamard) multiplication, and queries are scored by cosine similarity. Training combines three objectives: an InfoNCE-style contrastive loss with axis-aware hard negatives and a momentum queue, an HSIC term that pushes the two spaces toward independence, and a Jacobian penalty on cross-axis sensitivity. On Hindi, Bengali, Assamese, Sanskrit, Tamil and Telugu, the model reaches 81.12% accuracy on new languages with familiar labels and 63.5% on new labels in familiar languages, outperforming IndicBERT, mBERT and XLM-R. This lets NLU systems grow along either axis without retraining the encoder.

## Key points
- **Two evaluation regimes:**
  - Few-shot languages with known semantics: Assamese, Sanskrit, Tamil and Telugu are unseen.
  - Known languages with few-shot semantics: Hindi and Bengali.
- **Few-shot language setting (accuracy):**
  - Assamese 0.813, Sanskrit 0.747, Tamil 0.873, Telugu 0.880.
  - Gains over IndicBERT, the strongest baseline, range from +3.8 to +12.3 pp.
  - Language-identification F1 stays above 0.97 in all four languages.
- **Few-shot semantics setting:**
  - Accuracy is 0.645 for Hindi and 0.626 for Bengali.
  - IndicBERT scores 0.439 and 0.557 on the same splits.
- **Known languages with known semantics:**
  - Overall accuracy is 89.56%, slightly above IndicBERT (88.46%).
  - Macro F1 is a little lower than IndicBERT's (0.761 vs 0.779).
- **Ablation (joint accuracy):**
  - The full model scores 0.8956.
  - Dropping HSIC gives 0.886, dropping the Jacobian loss gives 0.879, and contrastive-only training gives 0.870.
- **Data source:** the extracted text omits the dataset section. The paper cites MASSIVE, and its intent labels follow MASSIVE's scenario/intent scheme.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [University of Birmingham](https://md.research.elamur.ai/free/en/universities/university-of-birmingham/) — University, Birmingham, United Kingdom · 27 works · 62,564 citations
- [Birla Institute of Technology and Science, Pilani](https://md.research.elamur.ai/free/en/universities/birla-institute-of-technology-and-science-pilani/) — University, Pilāni, India · 16 works · 41,879 citations
- [Birla Institute of Technology and Science - Hyderabad Campus](https://md.research.elamur.ai/free/en/universities/birla-institute-of-technology-and-science-hyderabad-campus/) — University, Hyderabad, India · 2 works

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works

🧭 Topics (2): [🌍 Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · 882 works · [🌱 Low-resource languages](https://md.research.elamur.ai/free/en/topics/low-resource-languages/) · 287 works

## Cite

Cite as: Kathakali Mitra, Sakshi Singh, Sree Nithish Reddy Gunapati, Aruna Malapati and Mark G. Lee. 2026. Dual-Axis Compositional Contrastive Few-Shot Learning using Prototypes Across Linguistic and Semantic Dimensions for Indic Low-Resource Multilingual NLU. Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{mitra2026dual,
  title = {{Dual-Axis Compositional Contrastive Few-Shot Learning using Prototypes Across Linguistic and Semantic Dimensions for Indic Low-Resource Multilingual NLU}},
  author = {Kathakali Mitra, Sakshi Singh, Sree Nithish Reddy Gunapati, Aruna Malapati and Mark G. Lee},
  booktitle = {Proceedings of the Sixth Workshop on Language Technology for Equality, Diversity, Inclusion},
  year = {2026},
  doi = {10.18653/v1/2026.ltedi-1.3},
  url = {https://doi.org/10.18653/v1/2026.ltedi-1.3}
}
```

- [← previous Dual Activation-Weight Sparsity: A Training-Free Framework for Efficient Large Language Model Compression](https://md.research.elamur.ai/free/en/papers/sun-2026-dual-activation-weight-sparsity-training/)
- [next → Dual-Cluster Memory Agent: Resolving Multi-Paradigm Ambiguity in Optimization Problem Solving](https://md.research.elamur.ai/free/en/papers/zhang-2026-dual-cluster-memory-agent-resolving/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/30/)

## References

19 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Junyi He and Xia Li. 2024. Zero-shot cross-lingual automated essay scoring. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 17819–17832. [find](https://scholar.google.com/scholar?q=Junyi+He+and+Xia+Li%2E+2024%2E+Zero%2Dshot+cross%2Dlingual+automated+essay+scoring%2E+In+Proceedings+of+the+2024+Joint+International+Conference+on+Computational+Linguistics%2C+Language+Resources+and+Evaluation+%28L)
2. Sanchit Ahuja, Divyanshu Aggarwal, Varun Gumma, Ishaan Watts, Ashutosh Sathe, Millicent Ochieng, Rishav Hada, Prachi Jain, Mohamed Ahmed, Kalika Bali, and Sunayana Sitaram. 2024. [MEGAVERSE: Benchmarking large language models across languages, modalities, models and tasks](https://md.research.elamur.ai/free/en/papers/ahuja-2024-megaverse-benchmarking-large-language-models/). In Proceedings of the 2024…
3. Pratik Joshi, Sebastin Santy, Amar Budhiraja, Kalika Bali, and Monojit Choudhury. 2020. [The state and fate of linguistic diversity and inclusion in the nlp world](https://md.research.elamur.ai/free/en/papers/joshi-2020-state-fate-linguistic-diversity-inclusion/). In Proceedings of the 58th annual meeting of the association for computational linguistics, pages 6282–6293.
4. Pei Cao, Yu Li, and Xinlu Li. 2025. Cross-language few-shot intent recognition via prompt-based tuning: P. cao et al. Applied Intelligence, 55(1):60. [find](https://scholar.google.com/scholar?q=Pei+Cao%2C+Yu+Li%2C+and+Xinlu+Li%2E+2025%2E+Cross%2Dlanguage+few%2Dshot+intent+recognition+via+prompt%2Dbased+tuning%3A+P%2E+cao+et+al%2E+Applied+Intelligence%2C+55%281%29%3A60%2E)
5. Divyanshu Kakwani, Anoop Kunchukuttan, Satish Golla, Gokul NC, Avik Bhattacharyya, Mitesh M Khapra, and Pratyush Kumar. 2020. [Indicnlpsuite: Monolingual corpora, evaluation benchmarks and pre-trained multilingual language models for indian languages](https://md.research.elamur.ai/free/en/papers/kakwani-2020-indicnlpsuite-monolingual-corpora-evaluation-benchmarks/). In Findings of the association for computational…
6. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. [Unsupervised cross-lingual representation learning at scale](https://md.research.elamur.ai/free/en/papers/conneau-2020-unsupervised-cross-lingual-representation-learning/). In Proceedings of the 58th annual meeting of the association fo…
7. Prannay Khosla, Piotr Teterwak, Chen Wang, Aaron Sarna, Yonglong Tian, Phillip Isola, Aaron Maschinot, Ce Liu, and Dilip Krishnan. 2020. Supervised contrastive learning. Advances in neural information processing systems, 33:18661–18673. [find](https://scholar.google.com/scholar?q=Prannay+Khosla%2C+Piotr+Teterwak%2C+Chen+Wang%2C+Aaron+Sarna%2C+Yonglong+Tian%2C+Phillip+Isola%2C+Aaron+Maschinot%2C+Ce+Liu%2C+and+Dilip+Krishnan%2E+2020%2E+Supervised+contrastive+learning%2E+Advances+in+neural+information)
8. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies,…
9. Aaron van den Oord, Yazhe Li, and Oriol Vinyals. 2018. [Representation learning with contrastive predictive coding](https://md.research.elamur.ai/free/en/papers/oord-2018-representation-learning-contrastive-predictive-coding/). arXiv preprint arXiv:1807.03748.
10. Soham Parikh, Mitul Tiwari, Prashil Tumbade, and Quaizar Vohra. 2023. Exploring zero and few-shot techniques for intent classification. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track), pages 744–751, Toronto, Canada. Association f… [find](https://scholar.google.com/scholar?q=Soham+Parikh%2C+Mitul+Tiwari%2C+Prashil+Tumbade%2C+and+Quaizar+Vohra%2E+2023%2E+Exploring+zero+and+few%2Dshot+techniques+for+intent+classification%2E+In+Proceedings+of+the+61st+Annual+Meeting+of+the+Association+for)
11. Sumanth Doddapaneni, Rahul Aralikatte, Gowtham Ramesh, Shreya Goyal, Mitesh M Khapra, Anoop Kunchukuttan, and Pratyush Kumar. 2023. [Towards leaving no indic language behind: Building monolingual corpora, benchmark and models for indic languages](https://md.research.elamur.ai/free/en/papers/doddapaneni-2023-leaving-no-indic-language-behind/). In Proceedings of the 61st Annual Meeting of the Assoc…
12. Telmo Pires, Eva Schlinger, and Dan Garrette. 2019. [How multilingual is multilingual bert?](https://md.research.elamur.ai/free/en/papers/pires-2019-how-multilingual-is-multilingual-bert/) In Proceedings of the 57th annual meeting of the association for computational linguistics, pages 4996–5001.
13. Guanting Dong, Zechen Wang, Liwen Wang, Daichi Guo, Dayuan Fu, Yuxiang Wu, Chen Zeng, Xuefeng Li, Tingfeng Hui, Keqing He, and 1 others. 2023. A prototypical semantic decoupling method via joint contrastive learning for few-shot named entity recognition. In ICASSP 2023-2023 IEEE International Confer… [find](https://scholar.google.com/scholar?q=Guanting+Dong%2C+Zechen+Wang%2C+Liwen+Wang%2C+Daichi+Guo%2C+Dayuan+Fu%2C+Yuxiang+Wu%2C+Chen+Zeng%2C+Xuefeng+Li%2C+Tingfeng+Hui%2C+Keqing+He%2C+and+1+others%2E+2023%2E+A+prototypical+semantic+decoupling+method+via+joint+contr)
14. Harman Singh, Nitish Gupta, Shikhar Bharadwaj, Dinesh Tewari, and Partha Talukdar. 2024. IndicGen-Bench: A multilingual benchmark to evaluate generation capabilities of LLMs on Indic languages. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long… [find](https://scholar.google.com/scholar?q=Harman+Singh%2C+Nitish+Gupta%2C+Shikhar+Bharadwaj%2C+Dinesh+Tewari%2C+and+Partha+Talukdar%2E+2024%2E+IndicGen%2DBench%3A+A+multilingual+benchmark+to+evaluate+generation+capabilities+of+LLMs+on+Indic+languages%2E+In+Pro)
15. Jake Snell, Kevin Swersky, and Richard Zemel. 2017. Prototypical networks for few-shot learning. Advances in neural information processing systems, 30. [find](https://scholar.google.com/scholar?q=Jake+Snell%2C+Kevin+Swersky%2C+and+Richard+Zemel%2E+2017%2E+Prototypical+networks+for+few%2Dshot+learning%2E+Advances+in+neural+information+processing+systems%2C+30%2E)
16. Zeyin Song, Yifan Zhao, Yujun Shi, Peixi Peng, Li Yuan, and Yonghong Tian. 2023. Learning with fantasy: Semantic-aware virtual contrastive constraint for few-shot class-incremental learning. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 24183–24192. [find](https://scholar.google.com/scholar?q=Zeyin+Song%2C+Yifan+Zhao%2C+Yujun+Shi%2C+Peixi+Peng%2C+Li+Yuan%2C+and+Yonghong+Tian%2E+2023%2E+Learning+with+fantasy%3A+Semantic%2Daware+virtual+contrastive+constraint+for+few%2Dshot+class%2Dincremental+learning%2E+In+Procee)
17. Oriol Vinyals, Charles Blundell, Timothy Lillicrap, Daan Wierstra, and 1 others. 2016. [Matching networks for one shot learning](https://md.research.elamur.ai/free/en/papers/vinyals-2016-matching-networks-one-shot-learning/). Advances in neural information processing systems, 29.
18. Genta Indra Winata, Liang-Kang Huang, Soumya Vadlamannati, and Yash Chandarana. 2023. Multilingual few-shot learning via language model retrieval. arXiv preprint arXiv:2306.10964. — [arXiv:2306.10964](https://arxiv.org/abs/2306.10964) [find](https://scholar.google.com/scholar?q=Genta+Indra+Winata%2C+Liang%2DKang+Huang%2C+Soumya+Vadlamannati%2C+and+Yash+Chandarana%2E+2023%2E+Multilingual+few%2Dshot+learning+via+language+model+retrieval%2E+arXiv+preprint+arXiv%3A2306%2E10964%2E+%E2%80%94+%5BarXiv%3A2306%2E10964%5D)
19. Xiaotong Zhang, Xinyi Li, Feng Zhang, Zhiyi Wei, Junfeng Liu, and Han Liu. 2024. A coarse-to-fine prototype learning approach for multi-label few-shot intent detection. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 2489– 2502, Miami, Florida, USA. Association for Co… [find](https://scholar.google.com/scholar?q=Xiaotong+Zhang%2C+Xinyi+Li%2C+Feng+Zhang%2C+Zhiyi+Wei%2C+Junfeng+Liu%2C+and+Han+Liu%2E+2024%2E+A+coarse%2Dto%2Dfine+prototype+learning+approach+for+multi%2Dlabel+few%2Dshot+intent+detection%2E+In+Findings+of+the+Association+)

*Updated 2 November 2026*