no code implementations • 7 Mar 2025 • Dong Shu, Xuansheng Wu, Haiyan Zhao, Daking Rai, Ziyu Yao, Ninghao Liu, Mengnan Du
Despite the challenges that remain around SAE implementation and scaling, they continue to provide valuable tools for understanding the internal mechanisms of large language models.
no code implementations • 21 Feb 2025 • Xuansheng Wu, Jiayi Yuan, Wenlin Yao, Xiaoming Zhai, Ninghao Liu
Large language models (LLMs) excel at handling human queries, but they can occasionally generate flawed or unexpected responses.
no code implementations • 19 Feb 2025 • Xuansheng Wu, Wenhao Yu, Xiaoming Zhai, Ninghao Liu
In training the classification model, we propose a simple and effective regularizer, by minimizing the similarity between the classifier weights and the identified unintended feature, to remove the impacts of these unintended features toward classification.
no code implementations • 2 Oct 2024 • Zhenyue Qin, Yu Yin, Dylan Campbell, Xuansheng Wu, Ke Zou, Yih-Chung Tham, Ninghao Liu, Xiuzhen Zhang, Qingyu Chen
The prevalence of vision-threatening eye diseases is a significant global burden, with many cases remaining undiagnosed or diagnosed too late for effective treatment.
no code implementations • 4 Jul 2024 • Xuansheng Wu, Padmaja Pravin Saraf, Gyeonggeon Lee, Ehsan Latif, Ninghao Liu, Xiaoming Zhai
Specifically, we prompt LLMs to generate analytic rubrics that they use to assign scores and study the alignment gap with human grading rubrics.
1 code implementation • 28 Mar 2024 • Yucheng Shi, Qiaoyu Tan, Xuansheng Wu, Shaochen Zhong, Kaixiong Zhou, Ninghao Liu
To tackle the problem, we propose the Retrieval-Augmented model Editing (RAE) framework for multi-hop question answering.
1 code implementation • 13 Mar 2024 • Xuansheng Wu, Haiyan Zhao, Yaochen Zhu, Yucheng Shi, Fan Yang, Tianming Liu, Xiaoming Zhai, Wenlin Yao, Jundong Li, Mengnan Du, Ninghao Liu
Therefore, in this paper, we introduce Usable XAI in the context of LLMs by analyzing (1) how XAI can benefit LLMs and AI systems, and (2) how LLMs can contribute to the advancement of XAI.
1 code implementation • 7 Jan 2024 • Yiwei Qin, Kaiqiang Song, Yebowen Hu, Wenlin Yao, Sangwoo Cho, Xiaoyang Wang, Xuansheng Wu, Fei Liu, PengFei Liu, Dong Yu
This paper introduces the Decomposed Requirements Following Ratio (DRFR), a new metric for evaluating Large Language Models' (LLMs) ability to follow instructions.
no code implementations • 30 Nov 2023 • Gyeong-Geon Lee, Ehsan Latif, Xuansheng Wu, Ninghao Liu, Xiaoming Zhai
We found a more balanced accuracy across different proficiency categories when CoT was used with a scoring rubric, highlighting the importance of domain-specific reasoning in enhancing the effectiveness of LLMs in scoring tasks.
1 code implementation • 30 Sep 2023 • Xuansheng Wu, Wenlin Yao, Jianshu Chen, Xiaoman Pan, Xiaoyang Wang, Ninghao Liu, Dong Yu
In this work, we investigate how the instruction tuning adjusts pre-trained models with a focus on intrinsic changes.
1 code implementation • 29 Jun 2023 • Xuansheng Wu, Huachi Zhou, Yucheng Shi, Wenlin Yao, Xiao Huang, Ninghao Liu
To evaluate our approach, we introduce a cold-start recommendation benchmark, and the results demonstrate that the enhanced small language models can achieve comparable cold-start recommendation performance to that of large models with only $17\%$ of the inference time.
no code implementations • 24 Apr 2023 • Ehsan Latif, Gengchen Mai, Matthew Nyaaba, Xuansheng Wu, Ninghao Liu, Guoyu Lu, Sheng Li, Tianming Liu, Xiaoming Zhai
AGI, driven by the recent large pre-trained models, represents a significant leap in the capability of machines to perform tasks that require human-level intelligence, such as reasoning, problem-solving, decision-making, and even understanding human emotions and social interactions.
1 code implementation • NeurIPS 2023 • Yucheng Shi, Mengnan Du, Xuansheng Wu, Zihan Guan, Jin Sun, Ninghao Liu
Defending against such attacks is challenging, especially for real-world black-box models where only query access is permitted.
no code implementations • 13 Mar 2023 • Xuansheng Wu, Kaixiong Zhou, Mingchen Sun, Xin Wang, Ninghao Liu
In particular, we introduce the basic concepts of graph prompt learning, organize the existing work of designing graph prompting functions, and describe their applications and future challenges.
1 code implementation • 24 Feb 2023 • Xuansheng Wu, Zhiyi Zhao, Ninghao Liu
We propose a novel non-parametric/un-trainable language model, named Non-Parametric Pairwise Attention Random Walk Model (NoPPA), to generate sentence embedding only with pre-trained word embedding and pre-counted word frequency.
1 code implementation • 20 Jan 2023 • Xuansheng Wu, Xinyu He, Tianming Liu, Ninghao Liu, Xiaoming Zhai
Developing models to automatically score students' written responses to science problems is critical for science education.