Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unsupervised Domain Adaptation on Reading Comprehension
Yu Cao, Meng Fang, Baosheng Yu, Joey Tianyi Zhou
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 |
|---|---|---|---|
| apply CASe adaptation instead of direct zero-shot transfer | BERT base-uncased, CNN→SQuAD · SQuAD dev | EM: 18.97 -> 65.24 | Improved |
| apply CASe adaptation instead of direct zero-shot transfer | BERT base-uncased, DailyMail→CNN · CNN dev | EM: 77.22 -> 82.44 | Improved |
| apply CASe adaptation instead of direct zero-shot transfer | BERT base-uncased, CNN→NewsQA · NewsQA dev | EM: 9.38 -> 48.95 | Improved |
| apply CASe adaptation instead of direct zero-shot transfer | BERT base-uncased, SQuAD→CoQA · CoQA dev | EM: 29.58 -> 50.77 | Improved |
| add conditional discriminator conditioning on outputs (ablation) | CASe, CNN→SQuAD · SQuAD dev | EM: 64.47 -> 65.24 | Improved |
| add conditional discriminator conditioning on outputs (ablation) | CASe, DailyMail→CNN · CNN dev | EM: 82.26 -> 82.44 | Improved |
| add conditional discriminator conditioning on outputs (ablation) | CASe, CNN→NewsQA · NewsQA dev | EM: 47.31 -> 48.95 | Improved |
| add conditional discriminator conditioning on outputs (ablation) | CASe, SQuAD→CoQA · CoQA dev | EM: 50.25 -> 50.77 | Improved |
| add adversarial learning (ablation) | CASe, CNN→SQuAD · SQuAD dev | EM: 65.05 -> 65.24 | Improved |
| add adversarial learning (ablation) | CASe, DailyMail→CNN · CNN dev | EM: 81.21 -> 82.44 | Improved |
| add adversarial learning (ablation) | CASe, CNN→NewsQA · NewsQA dev | EM: 47.89 -> 48.95 | Improved |
| add adversarial learning (ablation) | CASe, SQuAD→CoQA · CoQA dev | EM: 49.05 -> 50.77 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- Self — SELF: Learning to Filter Noisy Labels with Self-Ensembling (ICLR 2020) · better on 1 of 4 shared results
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
- Adapting Large Language Models via Reading Comprehension (ICLR 2024 · similarity 0.87)
- A Robust Adversarial Training Approach to Machine Reading Comprehension (AAAI 2020 · similarity 0.87)
- Domain-Adversarial Training of Neural Networks (JMLR 2016 · similarity 0.86)
- Robust Domain Adaptation for Machine Reading Comprehension (AAAI 2023 · similarity 0.86)
- Towards Robust Knowledge Unlearning: An Adversarial Framework for Assessing and Improving Unlearning Robustness in Large Language Models (AAAI 2025 · similarity 0.85)
- Auto-GDA: Automatic Domain Adaptation for Efficient Grounding Verification in Retrieval-Augmented Generation (ICLR 2025 · 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.
