Zobrazeno 1 - 10
of 192
pro vyhledávání: '"Jin, Lifeng"'
Despite the impressive capabilities of Large Language Models (LLMs) on various tasks, they still struggle with scenarios that involves complex reasoning and planning. Recent work proposed advanced prompting techniques and the necessity of fine-tuning
Externí odkaz:
http://arxiv.org/abs/2404.12253
Large language models (LLMs) exhibit impressive natural language capabilities but suffer from hallucination -- generating content ungrounded in the realities of training data. Recent work has focused on decoding techniques to improve factuality durin
Externí odkaz:
http://arxiv.org/abs/2404.09338
Autor:
Wang, Ante, Song, Linfeng, Tian, Ye, Peng, Baolin, Jin, Lifeng, Mi, Haitao, Su, Jinsong, Yu, Dong
Calibration, which establishes the correlation between accuracy and model confidence, is important for LLM development. We design three off-the-shelf calibration methods based on self-consistency (Wang et al., 2022) for math reasoning tasks. Evaluati
Externí odkaz:
http://arxiv.org/abs/2403.09849
Knowledge-based, open-domain dialogue generation aims to build chit-chat systems that talk to humans using mined support knowledge. Many types and sources of knowledge have previously been shown to be useful as support knowledge. Even in the era of l
Externí odkaz:
http://arxiv.org/abs/2403.03496
The most common training pipeline for large language models includes pretraining, finetuning and aligning phases, with their respective resulting models, such as the pretrained model and the finetuned model. Finetuned and aligned models show improved
Externí odkaz:
http://arxiv.org/abs/2402.17982
Autor:
Wang, Ante, Song, Linfeng, Peng, Baolin, Tian, Ye, Jin, Lifeng, Mi, Haitao, Su, Jinsong, Yu, Dong
This work studies improving large language model (LLM) generations at inference time by mitigating fact-conflicting hallucinations. Particularly, we propose a self-endorsement framework that leverages the fine-grained fact-level comparisons across mu
Externí odkaz:
http://arxiv.org/abs/2402.15631
Autor:
Zhang, Xiaoying, Peng, Baolin, Tian, Ye, Zhou, Jingyan, Jin, Lifeng, Song, Linfeng, Mi, Haitao, Meng, Helen
Publikováno v:
ACL2024 Main
Despite showing increasingly human-like abilities, large language models (LLMs) often struggle with factual inaccuracies, i.e. "hallucinations", even when they hold relevant knowledge. To address these hallucinations, current approaches typically nec
Externí odkaz:
http://arxiv.org/abs/2402.09267
One critical issue for chat systems is to stay consistent about preferences, opinions, beliefs and facts of itself, which has been shown a difficult problem. In this work, we study methods to assess and bolster utterance consistency of chat systems.
Externí odkaz:
http://arxiv.org/abs/2401.10353
Autor:
Xie, Shuyi, Yao, Wenlin, Dai, Yong, Wang, Shaobo, Zhou, Donlin, Jin, Lifeng, Feng, Xinhua, Wei, Pengzhi, Lin, Yujie, Hu, Zhichao, Yu, Dong, Zhang, Zhengyou, Nie, Jing, Liu, Yuhong
Large language models (LLMs) have shown impressive capabilities across various natural language tasks. However, evaluating their alignment with human preferences remains a challenge. To this end, we propose a comprehensive human evaluation framework
Externí odkaz:
http://arxiv.org/abs/2311.05374
Autor:
Shen, Lingfeng, Chen, Sihao, Song, Linfeng, Jin, Lifeng, Peng, Baolin, Mi, Haitao, Khashabi, Daniel, Yu, Dong
Standard practice within Reinforcement Learning from Human Feedback (RLHF) involves optimizing against a Reward Model (RM), which itself is trained to reflect human preferences for desirable generations. A notable subject that is understudied is the
Externí odkaz:
http://arxiv.org/abs/2309.16155