LLM

PIS꞉ Linking Importance Sampling and Attention Mechanisms for Efficient Prompt Compression
Large language models (LLMs) have achieved remarkable progress, demonstrating unprecedented capabilities across various natural …
Lizhe Chen, Binjia Zhou, Yuyao Ge 葛钰峣, Jiayi Chen, Shiguang NI
PDF Cite DOI
PIS꞉ Linking Importance Sampling and Attention Mechanisms for Efficient Prompt Compression
a1꞉ Steep Test-time Scaling Law via Environment Augmented Generation
Large Language Models (LLMs) have made remarkable breakthroughs in reasoning, yet continue to struggle with hallucinations, logical …
Lingrui Mei, Shenghua Liu, Yiwei Wang, Baolong Bi, Yuyao Ge 葛钰峣, Jun Wan, Yurong Wu, Xueqi Cheng
PDF Cite DOI
a1꞉ Steep Test-time Scaling Law via Environment Augmented Generation
Innate Reasoning is Not Enough : In-Context Learning Enhances Reasoning Large Language Models with Less Overthinking
TBD
Yuyao Ge 葛钰峣, Shenghua Liu, Yiwei Wang, Lingrui Mei, Lizhe Chen, Baolong Bi, Xueqi Cheng
PDF Cite arXiv
Innate Reasoning is Not Enough : In-Context Learning Enhances Reasoning Large Language Models with Less Overthinking
EMNLP2024论文分享 | Fewer is More：CoT示例要少而精
作者提出CoT-Influx方法，一种对CoT的示例和内容进行优化从而提高LLMs推理能力的方法，其核心思想是通过剪枝最大化有效信息的输入。
Yuyao Ge 葛钰峣
Oct 24, 2024 2 min read 论文分享
EMNLP2024论文分享 | Fewer is More：CoT示例要少而精
Can Graph Descriptive Order Affect Solving Graph Problems with LLMs?
In this work, we propose a Dynamically Adaptive Density Control Strategy based on the degree of reconstruction of the background of the scene, which adaptive the spatial sample point generation strategy dynamically according to the training results and prevents the generation of redundant data in the model.
Yuyao Ge 葛钰峣, Shenghua Liu, Baolong Bi, Yiwei Wang, Lingrui Mei, Wenjie Feng, Lizhe Chen, Xueqi Cheng
PDF Cite Project arXiv
Can Graph Descriptive Order Affect Solving Graph Problems with LLMs?
Translating Words to Worlds: Zero-Shot Synthesis of 3D Terrain from Textual Descriptions Using Large Language Models
The current research on text-guided 3D synthesis predominantly utilizes complex diffusion models, posing significant challenges in …
Guangzi Zhang, Lizhe Chen, Yu Zhang, Yan Liu, Yuyao Ge 葛钰峣, Xingquan Cai
PDF Cite DOI
Translating Words to Worlds: Zero-Shot Synthesis of 3D Terrain from Textual Descriptions Using Large Language Models
论文分享 | 广泛的解码策略导致大模型越狱
在本文，作者提出了一个新的数据集MaliciousInstruct，一种模型回答毒性评估方式，一种通过操纵解码超参数的攻击手段——generation exploitation，一种对齐策略——generation-aware alignment
Yuyao Ge 葛钰峣
Apr 9, 2024 1 min read 论文分享
论文分享 | 广泛的解码策略导致大模型越狱
论文解读 | TTA:大模型回答置信度评估新方法
本文提出了一种新的方法，全面评估大模型多个候选答案的可信度，以减轻大模型对于错误答案的过度自信。
Yuyao Ge 葛钰峣
Mar 25, 2024 2 min read 论文分享
论文解读 | TTA:大模型回答置信度评估新方法
Softmax回归及其优化问题
本文所属系列为笔者学习陈天奇和J.Zico Kolter在CMU开设的Deep Learning Systems的课程笔记。
Yuyao Ge 葛钰峣
Mar 21, 2024 3 min read 笔记
Softmax回归及其优化问题
论文解读 | 3月最新用于游戏的大模型Agent综述
3月最新用于游戏的大模型Agent综述
Yuyao Ge 葛钰峣
Mar 21, 2024 1 min read 论文分享
论文解读 | 3月最新用于游戏的大模型Agent综述