Research map / Label-efficient and robust learning / Recognition and long-tail
HC2L: Hybrid and Cooperative Contrastive Learning for Cross-Lingual Spoken Language Understanding
Bowen Xing, Ivor Wai-Hung Tsang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add supervised contrastive learning (source, cross-lingual, multilingual) on top of cross-lingual unsupervised CL | GL-CLEF (mBERT) · MultiATIS++ | overall accuracy: 53.56 -> 58.98 | Improved |
| add supervised CL to unsupervised CL (UCL+SCL vs only UCL) (ablation) | HC2L · MultiATIS++ | all metrics | Improved |
| use only supervised CL instead of only unsupervised CL (ablation) | HC2L · MultiATIS++ | all metrics | Improved |
| add source language / cross-lingual / multilingual supervised CL individually (ablation) | HC2L · MultiATIS++ | all metrics | Improved |
| add joint-task multi-label supervised CL (ablation) | HC2L · MultiATIS++ | all metrics | Improved |
| add intent SCL and slot SCL single-task mechanisms (ablation) | HC2L · MultiATIS++ | all metrics | Improved |
| swap encoder to XLM-R and add HC2L | XLM-R · MultiATIS++ (XLM-R) | overall accuracy: 33.31 -> 56.06 | Improved |
| replace GL-CLEF with HC2L on XLM-R | XLM-R + GL-CLEF · MultiATIS++ (XLM-R) | overall accuracy: 52.69 -> 56.06 | Improved |
| replace GL-CLEF with HC2L on XLM-R | XLM-R + GL-CLEF · MultiATIS++ (XLM-R) | intent accuracy: 92.75 -> 94.85 | Improved |
| use HC2L on Hindi | GL-CLEF · MultiATIS++ Hindi | overall accuracy: 26.76 -> 34.83 | Improved |
| increase supervised CL weights beyond tuned peak or reduce them (ablation) | HC2L · MultiATIS++ | overall accuracy | Hurt |
| HC2L with smaller queue (16 vs 32) and extra losses | GL-CLEF · computation | GPU memory / training time: 20.7GB -> 18.9GB; 33s -> 35s | Mixed |
Closest papers by meaning
- DiCA: Disambiguated Contrastive Alignment for Cross-Modal Retrieval with Partial Labels (AAAI 2025 · similarity 0.85)
- Soft Token Alignment for Cross-Lingual Reasoning (NeurIPS 2026 · similarity 0.84)
- Constrained Decoding for Cross-lingual Label Projection (ICLR 2024 · similarity 0.84)
- MCL-NER: Cross-Lingual Named Entity Recognition via Multi-View Contrastive Learning (AAAI 2024 · similarity 0.84)
- Dynamic Adapter with Semantics Disentangling for Cross-lingual Cross-modal Retrieval (AAAI 2025 · similarity 0.84)
- CL2CM: Improving Cross-Lingual Cross-Modal Retrieval via Cross-Lingual Knowledge Transfer (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
