Research map / Label-efficient and robust learning / Domain adaptation and generalization
Learning Beyond Domains: Misleading Prompts and Pseudo-Label Contrast for Text Domain Generalization
Qizhi Li, Xuyang Wang, Yingke Chen, Ming Yan, Dezhong Peng, Xi Peng, Xu Wang
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 domain-misleading prompt learning (DMPL) (ablation) | GenPromptCL · Amazon (DEK->B) | macro-F1: 91.13 -> 91.67 | Improved |
| add pseudo-label contrastive learning (PCL) (ablation) | GenPromptCL · Amazon (BDE->K) | macro-F1: 93.42 -> 94.03 | Improved |
| add both DMPL and PCL (ablation) | GenPromptCL · Amazon (BEK->D) | macro-F1: 89.45 -> 90.18 | Improved |
| add either DMPL or PCL alone on top of the base prompt classifier (ablation) | GenPromptCL · Amazon (BDE->K) | macro-F1: 93.47 -> 93.48 (DMPL), 93.42 (PCL) | No clear effect |
| GenPromptCL versus PDA adversarial training | PHEME rumor detection · PHEME (avg) | macro-F1: 40.77 -> 78.18 | Improved |
| GenPromptCL versus best baseline on Amazon | Amazon reviews · Amazon (avg) | macro-F1: 91.85 -> 92.04 | Improved |
| GenPromptCL versus best baseline on PHEME | PHEME · PHEME (avg) | macro-F1: 75.47 -> 78.18 | Improved |
| GenPromptCL versus best baseline on MNLI | MNLI · MNLI (avg) | macro-F1: 80.89 -> 81.74 | Improved |
| GenPromptCL versus SimCSE on Amazon->IMDB | cross-dataset transfer · Amazon->IMDB | macro-F1: 90.96 -> 90.77 | No clear effect |
| GenPromptCL versus best baseline in cross-dataset average | cross-dataset transfer · Cross-dataset avg | macro-F1: 77.26 -> 77.83 | Improved |
| replace adversarial/learned-parameter DG modules with prompt-based DMPL | domain generalization · Amazon (one domain) | additional parameters (M): 124.84 (MSCL) -> 0.20 | Improved |
| GenPromptCL versus MSCL on training speed | domain generalization · Amazon (one domain) | runtime per epoch (s): 51.74 -> 58.11 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- TACIT — TACIT: A Target-Agnostic Feature Disentanglement Framework for Cross-Domain Text Classification (AAAI 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Adversarial Domain Prompt Tuning and Generation for Single Domain Generalization (CVPR 2025 · similarity 0.86)
- Learning to Generate Novel Domains for Domain Generalization (ECCV 2020 · similarity 0.86)
- Concentrate Attention: Towards Domain-Generalizable Prompt Optimization for Language Models (NeurIPS 2024 · similarity 0.84)
- Gen-Z: Generative Zero-Shot Text Classification with Contextualized Label Descriptions (ICLR 2024 · similarity 0.84)
- Towards Generalizing to Unseen Domains with Few Labels (CVPR 2024 · similarity 0.84)
- Rethinking Multi-domain Generalization with A General Learning Objective (CVPR 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.
