| 1 |
CEMTM: Contextual Embedding-based Multimodal Topic Modeling |
CEMTM：基于上下文嵌入的多模态主题模型，提升短文档和长文档的主题一致性和可解释性。 |
multimodal |
|
|
| 2 |
Joint Effects of Argumentation Theory, Audio Modality and Data Enrichment on LLM-Based Fallacy Classification |
研究论证理论、音频模态与数据增强对LLM谬误分类的联合影响 |
large language model chain-of-thought |
|
|
| 3 |
Transformer Enhanced Relation Classification: A Comparative Analysis of Contextuality, Data Efficiency and Sequence Complexity |
对比Transformer与非Transformer模型在关系分类任务中的性能差异。 |
large language model |
|
|
| 4 |
Differentially-private text generation degrades output language quality |
差分隐私文本生成显著降低输出语言质量 |
large language model |
|
|
| 5 |
We Argue to Agree: Towards Personality-Driven Argumentation-Based Negotiation Dialogue Systems for Tourism |
提出PAN-DG任务与PACT数据集，用于旅游领域个性化论辩式协商对话生成。 |
large language model |
|
|
| 6 |
Continually Adding New Languages to Multilingual Language Models |
提出Layer-Selective LoRA，用于多语言模型持续添加新语言，避免灾难性遗忘。 |
instruction following |
|
|
| 7 |
The Prompt Engineering Report Distilled: Quick Start Guide for Life Sciences |
提炼Prompt工程报告，为生命科学领域提供LLM应用的快速入门指南 |
large language model |
|
|
| 8 |
Optimal Brain Restoration for Joint Quantization and Sparsification of LLMs |
提出Optimal Brain Restoration，解决LLM量化与稀疏化联合压缩难题 |
large language model |
|
|
| 9 |
Text2Mem: A Unified Memory Operation Language for Memory Operating System |
提出Text2Mem：一种统一的内存操作语言，用于提升Agent内存操作的可靠性和可移植性。 |
large language model |
|
|
| 10 |
When Smiley Turns Hostile: Interpreting How Emojis Trigger LLMs' Toxicity |
发现表情符号可触发大语言模型生成有害内容，并探究其内在机制 |
large language model |
|
|
| 11 |
Towards Better Health Conversations: The Benefits of Context-seeking |
提出Wayfinding AI，通过主动寻求上下文信息，改善LLM在健康咨询中的表现。 |
large language model |
|
|