Research map / Label-efficient and robust learning / Semi-supervised learning
Weakly-Supervised Hierarchical Models for Predicting Persuasive Strategies in Good-faith Textual Requests
Jiaao Chen, Diyi Yang
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 document-level supervision and document-level prediction to sentence VAE (ablation) | S-VAE · Kiva (all labeled training documents) | Macro F1: 55.7 ± 1.0 -> 55.4 ± 0.8 | No clear effect |
| Add document-level supervision and document-level prediction to sentence VAE (ablation) | S-VAE · RAOP (all labeled training documents) | Macro F1: 48.8 ± 2.0 -> 49.1 ± 2.0 | Improved |
| Add document-level supervision and document-level prediction to sentence VAE (ablation) | S-VAE · Borrow (all labeled training documents) | Macro F1: 74.6 ± 0.5 -> 74.7 ± 0.3 | No clear effect |
| Use BERT as the sentence encoder (ablation) | WS-VAE · Kiva (all labeled training documents) | Macro F1: 55.4 ± 0.8 -> 59.1 ± 0.9 | Improved |
| Use BERT as the sentence encoder (ablation) | WS-VAE · RAOP (all labeled training documents) | Macro F1: 49.1 ± 2.0 -> 58.3 ± 2.0 | Improved |
| Use BERT as the sentence encoder (ablation) | WS-VAE · Borrow (all labeled training documents) | Macro F1: 74.7 ± 0.3 -> 75.9 ± 0.7 | Improved |
| Set the KL threshold on y to 1.2 rather than 0 (ablation) | WS-VAE · Kiva (20 labeled documents) | Macro F1: 0.228 -> 0.315 | Improved |
| Set the KL threshold on y to 1.2 rather than 2.0 (ablation) | WS-VAE · Kiva (20 labeled documents) | Macro F1: 0.305 -> 0.315 | Improved |
| Set the KL threshold on y to 1.2 rather than 0 (ablation) | WS-VAE · RAOP (20 labeled documents) | Macro F1: 0.274 -> 0.321 | Improved |
| Set the KL threshold on y to 1.2 rather than 2.0 (ablation) | WS-VAE · RAOP (20 labeled documents) | Macro F1: 0.316 -> 0.321 | Improved |
| Set the KL threshold on y to 1.2 rather than 0 (ablation) | WS-VAE · Borrow (20 labeled documents) | Macro F1: 0.485 -> 0.595 | Improved |
| Set the KL threshold on y to 1.2 rather than 2.0 (ablation) | WS-VAE · Borrow (20 labeled documents) | Macro F1: 0.542 -> 0.595 | Improved |
Closest papers by meaning
- Measuring And Improving Persuasiveness Of Large Language Models (ICLR 2025 · similarity 0.87)
- Debating with More Persuasive LLMs Leads to More Truthful Answers (ICML 2024 · similarity 0.84)
- Maximizing Signal in Human-Model Preference Alignment (AAAI 2025 · similarity 0.84)
- Scalable Influence and Fact Tracing for Large Language Model Pretraining (ICLR 2025 · similarity 0.84)
- Debate or Vote: Which Yields Better Decisions in Multi-Agent Large Language Models? (NeurIPS 2025 · similarity 0.83)
- Bayesian WeakS-to-Strong from Text Classification to Generation (ICLR 2025 · 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.
