A5003425900- Fatigue and fracture mechanics
- Structural Load-Bearing Analysis
- Metal Forming Simulation Techniques
- Topic Modeling
- Natural Language Processing Techniques
- Fire effects on concrete materials
- High-Velocity Impact and Material Behavior
- Structural Response to Dynamic Loads
- Multimodal Machine Learning Applications
- Advanced Text Analysis Techniques
- Text and Document Classification Technologies
- Adversarial Robustness in Machine Learning
- Adaptive Dynamic Programming Control
- Reinforcement Learning in Robotics
Zhengzhou University
2022-2024
China Aerospace Science and Technology Corporation
2022
As the latest advancements in natural language processing, large models (LLMs) have achieved human-level understanding and generation abilities many real-world tasks, even been regarded as a potential path to artificial general intelligence. To better facilitate research on LLMs, open-source such Llama 2 Falcon, recently proposed gained comparable performances proprietary models. However, these are primarily designed for English scenarios exhibit poor Chinese contexts. In this technical...
Joint extraction from unstructured text aims to extract relational triples composed of entity pairs and their relations. However, most existing works fail process the overlapping issues that occur when same entities are utilized generate different in a sentence. In this work, we propose mutually exclusive Binary Cross Tagging (BCT) scheme develop end-to-end BCT framework jointly triples. Each token is assigned binary tag, then these tags cross-matched all tag sequences form Our method...
Abstract Reinforcement learning in a multi-agent setting is very important for real-world applications, but it brings more challenges than those single-agent environment. In the setting, agent generally has bias of overestimation on value function. our work, we pay attention to issue with continuous actions We propose method reduce this by adopting distributional perspective reinforcement learning. combine within framework off-policy Actor-Critic and novel approach Multi-Agent Deep...
In this paper, an abstractive text summarization method with document sharing is proposed. It consists of a pretrained model and self-attention mechanism on multi-document. We call it DoS mechanism. By applying the to single-document task, can absorb information from multiple documents, thus enhancing its effectiveness model. compared results several models. The experimental show that pre-trained modified attention provides best results, where values Rouge-l, Rouge-2, Rouge-L are 41.3%,...