Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Zero-Shot Adaptive Transfer for Conversational Language Understanding
Sungjin Lee, Rahul Jha
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 |
|---|---|---|---|
| Use slot-description-conditioned zero-shot adaptive transfer | Slot tagging across 10 domains · 10 domains, 2000 training instances | average F1 improvement over CRF: +7.24* | Improved |
| Use slot-description-conditioned zero-shot adaptive transfer | Slot tagging across 10 domains · 10 domains, 500 training instances | average F1 improvement over CRF: +14.67* | Improved |
| Use ZAT instead of Concept Tagger | Zero-shot slot tagging · 10 domains, 0 target training instances | average F1 improvement: +5.07 | Improved |
| Add a CRF layer (ablation) | ZAT · 10 domains, 0/500/1000/2000 target training instances | F1: 35.06 vs 36.05; 74.39 vs 76.04; 78.41 vs 80.26; 81.58 vs 82.35 | Improved |
| Add character-level word embeddings (ablation) | ZAT · 10 domains, 0/500/1000/2000 target training instances | F1: 35.49 vs 36.05; 73.71 vs 76.04; 77.86 vs 80.26; 81.11 vs 82.35 | Improved |
| Fine-tune pre-trained word embeddings (ablation) | ZAT · 10 domains, 0 target training instances | F1: 36.05 -> 33.71 | Hurt |
| Fine-tune pre-trained word embeddings (ablation) | ZAT · 10 domains, 500/1000/2000 target training instances | F1: 76.04 -> 76.52; 80.26 -> 80.61; 82.35 -> 83.09 | Improved |
| Use slot-description-conditioned ZAT rather than CT | Slot tagging · 10 domains, 1000 training instances | average F1 improvement over CT: +2.44 | Improved |
Closest papers by meaning
- Agent Instructs Large Language Models to be General Zero-Shot Reasoners (ICML 2024 · similarity 0.85)
- Exploring Conversational Adaptability: Assessing the Proficiency of Large Language Models in Dynamic Alignment with Updated User Intent (AAAI 2025 · similarity 0.84)
- SemTra: A Semantic Skill Translator for Cross-Domain Zero-Shot Policy Adaptation (AAAI 2024 · similarity 0.84)
- Model-Aware Tokenizer Transfer (NeurIPS 2026 · similarity 0.83)
- Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass (ICLR 2025 · similarity 0.83)
- Zero-Shot Continuous Prompt Transfer: Generalizing Task Semantics Across Language Models (ICLR 2024 · similarity 0.83)
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.
