Papers by Kaihe Xu
SA-DETR:Span Aware Detection Transformer for Moment Retrieval (2025.coling-main)
Copied to clipboard
| Challenge: | Moment Retrieval aims to locate video segments related to text. |
| Approach: | They propose a method that leverages the importance of instance related span anchors . they initialize span anchor using instance related fuse token and supervise them with GT labels . |
| Outcome: | The proposed method achieves competitive results on QVHighlights, Charades-STA and TACoS. |
Reinforcement Learning with Token-level Feedback for Controllable Text Generation (2024.findings-naacl)
Copied to clipboard
| Challenge: | Existing methods for controllable text generation are guided by coarse-grained feedback, which may lead to suboptimal performance owing to semantic twists or progressions within sentences. |
| Approach: | They propose a reinforcement learning algorithm which formulates TOken-LEvel rewards for controllable text generation and employs a "first-quantize-then-noise" paradigm to enhance the robustness of the RL algorithm. |
| Outcome: | The proposed algorithm can achieve superior performance on single-attribute and multi-attract control tasks. |
CoMIF: Modeling of Complex Multiple Interaction Factors for Conversation Generation (2025.coling-main)
Copied to clipboard
| Challenge: | Existing methods for generating human-like dialogues lack implicit correlations among factors . different factors may alternately dominate token-level response generation during decoding . |
| Approach: | They propose a framework that can model complex multiple interaction factors to generate human-like conversations. |
| Outcome: | The proposed framework generates human-like conversations with multiple factors compared to state-of-the-art methods . et al. show that the proposed framework produces superior results over existing methods compared with the state- of-the art methods based on multiple datasets . |
CNEQ: Incorporating numbers into Knowledge Graph Reasoning (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Complex query answering (CQA) is a task that addresses semantics of numerical entities. |
| Approach: | They propose a model that includes a Number-Entity Predictor and an Entity Filter . they use three widely-used Knowledge Graphs to perform reasoning over knowledge graphs . |
| Outcome: | The proposed model can predict entities and numerical values better than existing models . it compares or filters out entities that meet certain constraints on three widely-used Knowledge Graphs . |