• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.222-227
• /
• 2023

환각 현상이란 LLM이 생성 태스크에서 사실이 아닌 내용을 생성하거나 근거가 없는 내용을 생성하는 현상을 말한다. 환각 현상은 LLM이 생성한 출력물에 대한 사용자의 신뢰를 떨어뜨리기 때문에 환각을 완화할 수 있는 방법이 필요하다. 최근 사후 편집 모델 중 하나인 RARR는 입력 텍스트를 질문들 순서에 따라 순차적으로 편집하여 환각을 완화하였지만 이전 단계의 편집 오류가 전파되거나 같은 작업을 반복하는 등의 단점이 있었다. 본 논문은 환각 현상 완화를 위한 단위 사실 기반 사후 교정을 제안한다. 제안한 방법은 입력 텍스트를 단위 사실로 분해하고 각 사실에 대응하는 질문을 생성한 후 검색된 관련 문서로 환각 여부를 판단한다. 환각이라 판단되면 편집을 수행하여 환각을 완화한다. 병렬적으로 편집을 진행하기 때문에 기존 연구의 순차적인 오류 전파 문제를 해결하고 기존 연구에 비해 더 빠른 사후 편집을 진행할 수 있다. 실험 결과, 제안 방법이 RARR보다 Preservation Score, 원문과의 사실성 일치여부, 의도 보존 여부에서 모두 우수한 성능을 보인다.

• Informatization Policy
• /
• v.31 no.2
• /
• pp.3-38
• /
• 2024

Hallucination is a significant barrier to the utilization of large-scale language models or multimodal models. In this study, we collected 654 computer science papers with "hallucination" in the abstract from arXiv from December 2022 to January 2024 following the advent of Chat GPT and conducted frequency analysis, knowledge network analysis, and literature review to explore the latest trends in hallucination research. The results showed that research in the fields of "Computation and Language," "Artificial Intelligence," "Computer Vision and Pattern Recognition," and "Machine Learning" were active. We then analyzed the research trends in the four major fields by focusing on the main authors and dividing them into data, hallucination detection, and hallucination mitigation. The main research trends included hallucination mitigation through supervised fine-tuning (SFT) and reinforcement learning with human feedback (RLHF), inference enhancement via "chain of thought" (CoT), and growing interest in hallucination mitigation within the domain of multimodal AI. This study provides insights into the latest developments in hallucination research through a technology-oriented literature review. This study is expected to help subsequent research in both engineering and humanities and social sciences fields by understanding the latest trends in hallucination research.

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.690-693
• /
• 2023

언어 모델의 파라미터 수의 지속적인 증가로 100B 단위의 거대 언어모델 LLM(Large Language Model)을 구성 할 정도로 언어 모델의 크기는 증가 해 왔다. 이런 모델의 크기와 함께 성장한 다양한 Task의 작업 성능의 향상과 함께, 발전에는 환각(Hallucination) 및 윤리적 문제도 함께 떠오르고 있다. 이러한 문제 중 특히 환각 문제는 모델이 존재하지도 않는 정보를 실제 정보마냥 생성한다. 이러한 잘못된 정보 생성은 훌륭한 성능의 LLM에 신뢰성 문제를 야기한다. 환각 문제는 정보 검색을 통하여 입력 혹은 내부 표상을 증강하면 증상이 완화 되고 추가적으로 성능이 향상된다. 본 논문에서는 한국어 질의 응답에서 검색 증강을 통하여 모델의 개선점을 확인한다.

• Journal of Hospice and Palliative Care
• /
• v.22 no.2
• /
• pp.100-104
• /
• 2019

Opioids are important drugs for the management of severe cancer pain without a ceiling effect. However, opioid administration leads to dose-limiting complications including drowsiness, hallucinations, delirium, respiratory depression, cognitive impairment, seizure, myoclonus, and hyperalgesia. Opioid-induced hyperalgesia (OIH) is a paradoxical phenomenon as opioid exposure increases pain sensitivity. Reducing or stopping opioids, opioid rotation, or co-administration of N-methyl-D-aspartate (NMDA) antagonists have been suggested for the management of OIH. In this study, we report two clinical cases of successful management of OIH in cancer pain patients that were treated with opioids.

• Journal of Life Science
• /
• v.31 no.7
• /
• pp.677-685
• /
• 2021

Cannabis sativa L. belongs to the Cannabaceae family and is an annual herbaceous flowing plant. The plants can be classified into narcotic marijuana and nonnarcotic hemp. Different parts of C. sativa L. have been used as food, medicine, cosmetics, fiber and textile. However, the use of leaf, flower, and seed of C. sativa L was forbidden in Korea in January 1977 as a result of the Cannabis Control Act due to the narcotic properties. The plant's mature stems have limited uses for the production of fiber and sheets. Recently, various cannabinoids, terpenes and essential fatty acids were identified from C. sativa L., and their safety and useful bio-activities, such as neuroprotective, anti-inflammation, antithrombosis, antiepileptic, and antimicrobial activities, and the relief of pain, have been highlighted. Furthermore, the process of reduction of tetrahydrocannabinol, a representative narcotic compound, and the isolation of cannabidiol, a nonnarcotic active compound in C. sativa L., have been determined. These findings resulted in the legalization of C. sativa L. in Korea for medical use in December 2018 and the exclusion of C. sativa L. from the narcotic list of the UN Commission on Narcotics Drugs (UNCND) in December 2020. Therefore, developments of various high-value added products have commenced worldwide. Additionally, in 2021, the Korean government deregulated special zones based on hemp. In this study, the current status and the prospect of the hemp industry, as well as essential techniques for developing new hemp products, are provided for the activation of the Korea Green-Rush.