Research map / Label-efficient and robust learning / Knowledge distillation
Exploiting the Ground-Truth: An Adversarial Imitation Based Knowledge Distillation Approach for Event Detection
Jian Liu, Yubo Chen, Kang Liu
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 event-argument embeddings to the teacher (ablation) | Teacher + emb · ACE 2005 English golden | F1: 68.8 -> 74.2 | Improved |
| Add entity embeddings to the teacher (ablation) | Teacher + emb · ACE 2005 English golden | F1: 68.8 -> 70.3 | Improved |
| Add entity embeddings on top of argument embeddings (ablation) | Teacher + emb + agt · ACE 2005 English golden | F1: 74.2 -> 74.8 | Improved |
| Use adversarial imitation student instead of teacher on predicted tags | Student-Final · ACE 2005 English predicted | F1: 70.6 -> 71.2 | Improved |
| Use adversarial imitation student instead of ArgATT | Student-Final · ACE 2005 English predicted | F1: 70.7 -> 71.2 | Improved |
| Use adversarial imitation student instead of CNNED | Student-Final · ACE 2005 Chinese predicted | F1: 57.8 -> 58.5 | Improved |
| Use predicted instead of golden annotations at test time | Teacher · ACE 2005 English | F1: 74.8 -> 70.6 | Hurt |
| Use predicted instead of golden annotations at test time | Teacher · ACE 2005 Chinese | F1: 61.9 -> 57.3 | Hurt |
| Use heterogeneous attention teacher and CNN student (ablation) | Student-Final · ACE 2005 English predicted | F1: 71.2 -> 67.7 | Hurt |
| Use CNN for both teacher and student (ablation) | Student-Final · ACE 2005 English predicted | F1: 71.2 -> 70.8 | Hurt |
| Remove attention layer from student (ablation) | Student-Final · ACE 2005 English predicted | F1: 71.2 -> 68.6 | Hurt |
| Add the adversarial discriminator (vs none) (ablation) | Student-Final · ACE 2005 English predicted | F1: 69.2 -> 71.2 | Improved |
Closest papers by meaning
- Adversarial Moment-Matching Distillation of Large Language Models (NeurIPS 2024 · similarity 0.86)
- Scale Up Event Extraction Learning via Automatic Training Data Generation (AAAI 2018 · similarity 0.85)
- Adversarial Data Augmentation for Task-Specific Knowledge Distillation of Pre-trained Transformers (AAAI 2022 · similarity 0.85)
- Towards Robust Knowledge Unlearning: An Adversarial Framework for Assessing and Improving Unlearning Robustness in Large Language Models (AAAI 2025 · similarity 0.82)
- Adversarial Paraphrasing: A Universal Attack for Humanizing AI-Generated Text (NeurIPS 2025 · similarity 0.82)
- Is a Large Language Model a Good Annotator for Event Extraction? (AAAI 2024 · similarity 0.82)
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.
