Papers by Yuanchen Shi
Beyond Coarse Labels: Fine-Grained Problem Augmentation and Multi-Dimensional Feedback for Emotional Support Conversation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing ESC datasets often use coarse-grained problem categories, limiting models’ ability to address users’ complex, overlapping challenges. |
| Approach: | They propose a generalizable fine-grained problem enhancement method that augments problem types, user scenarios, and profiles, enabling the construction of richer and more diverse ESC corpora. |
| Outcome: | The proposed method improves both automatic and human evaluation metrics across different models. |
Danger Depends on the Mind: A Theory-of-Mind Grounded Dataset and Model for Context-Dependent Dangerous Speech (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for dangerous speech detection rely on binary labels that ignore who is speaking and in what mental state. |
| Approach: | They propose a context-dependent variant of dangerous speech detection by grounding it in Theory-of-Mind. |
| Outcome: | The proposed model outperforms proprietary and open-source models with significantly fewer parameters. |