• Clinical and Experimental Pediatrics
• /
• 제63권6호
• /
• pp.232-236
• /
• 2020

Background: Children today get access to smartphones at an early age. However, their ability to use mobile apps has not yet been studied in detail. Purpose: This study aimed to assess the ability of children aged 2-8 years to perform touchscreen gestures and follow prompting techniques, i.e., ways apps provide instructions on how to use them. Methods: We developed one mobile app to test the ability of children to perform various touchscreen gestures and another mobile app to test their ability to follow various prompting techniques. We used these apps in this study of 90 children in a kindergarten and a primary school in New Delhi in July 2019. We noted the touchscreen gestures that the children could perform and the most sophisticated prompting technique that they could follow. Results: Two- and 3-year-old children could not follow any prompting technique and only a minority (27%) could tap the touchscreen at an intended place. Four- to 6-year-old children could perform simple gestures like a tap and slide (57%) and follow instructions provided through animation (63%). Seven- and 8-year-old children could perform more sophisticated gestures like dragging and dropping (30%) and follow instructions provided in audio and video formats (34%). We observed a significant difference between the number of touchscreen gestures that the children could perform and the number of prompting techniques that they could follow (F=544.0407, P<0.05). No significant difference was observed in the performance of female versus male children (P>0.05). Conclusion: Children gradually learn to use mobile apps beginning at 2 years of age. They become comfortable performing single-finger gestures and following nontextual prompting techniques by 8 years of age. We recommend that these results be considered in the development of mobile apps for children.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2023년도 추계학술발표대회
• /
• pp.487-488
• /
• 2023

본 논문에서는 RLHF 기반의 오픈소스 LLM인 llama-2-13b model을 FinQA task에 적용하여 그 성능을 확인해 보았다. 이때, CoT, few-shot과 같은 다양한 prompting 기법들을 적용해보며 어떤 방법이 가장 효과적인지 비교했다. 그 결과, 한 번(total)에 task를 수행한 경우 few-shot 예시를 2개 사용했을 때보다 3개 사용했을 때, subtask로 나누어 수행한 경우 prompt로 답(simple)만 제시했을 때보다 CoT 형식으로 주었을 때, 각각 24.85%의 정확도로 가장 높은 성능을 보였다.

• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 2022년도 제34회 한글 및 한국어 정보처리 학술대회
• /
• pp.343-347
• /
• 2022

최근 자연어처리 분야에서는 BERT, RoBERTa, 그리고 BART와 같은 사전 학습된 언어 모델 (Pre-trained Language Models, PLM) 기반 미세 조정 학습을 통하여 여러 하위 과업에서 좋은 성능을 거두고 있다. 이는 사전 학습된 언어 모델 및 데이터 집합의 크기, 그리고 모델 구성의 중요성을 보여주며 대규모 사전 학습된 언어 모델이 각광받는 계기가 되었다. 하지만, 거대한 모델의 크기로 인하여 실제 산업에서 쉽게 쓰이기 힘들다는 단점이 명백히 존재함에 따라 최근 매개변수 효율적인 미세 조정 및 Few-Shot 학습 연구가 많은 주목을 받고 있다. 본 논문은 Prompt tuning, Prefix tuning와 프롬프트 기반 미세 조정 (Prompt-based fine-tuning)을 결합한 Few-Shot 학습 연구를 제안한다. 제안한 방법은 미세 조정 ←→ 사전 학습 간의 지식 격차를 줄일 뿐만 아니라 기존의 일반적인 미세 조정 기반 Few-Shot 학습 성능보다 크게 향상됨을 보인다.

• 한국수학교육학회지시리즈D:수학교육연구
• /
• 제23권2호
• /
• pp.63-92
• /
• 2020

Student engagement in high-level, cognitively demanding instruction is pivotal for student learning. However, many teachers are unable to maintain such instruction, especially in instances of non-responsive students. This case study of three middle school teachers explores prompts that aim to move classroom discussions past student silence. Prompt sequences were categorized into Progressing, Focusing, and Redirecting Actions, and then analyzed for maintenance of high levels of cognitive demand. Results indicate that specific prompt types are prone to either raise or diminish the cognitive demand of a discussion. While Focusing Actions afforded students opportunities to process information on a more meaningful level, Progressing Actions typically lowered cognitive demand in an effort to get through mathematics content or a specific method or procedure. Prompts that raise cognitive demand typically start out as procedural or concrete and progress to include students' thoughts or ideas about mathematical concepts. This study aims to discuss five specific implications on how teachers can use prompting techniques to effectively maintain cognitively challenging discourse through moments of student silence.

• 한국정보과학회 언어공학연구회:학술대회논문집(한글 및 한국어 정보처리)
• /
• 한국정보과학회언어공학연구회 2023년도 제35회 한글 및 한국어 정보처리 학술대회
• /
• pp.372-376
• /
• 2023

본 연구에서는 비디오 하이라이트 검출 및 장면 추출을 위한 경량화된 모델인 Visual Context Learner (VCL)을 제안한다. 기존 연구에서는 매개변수가 고정된 CLIP을 비롯한 여러 피쳐 추출기에 학습 가능한 DETR과 같은 트랜스포머를 이어붙여서 학습을 한다. 하지만 본 연구는 경량화된 구조로 하이라이트 검출 성능을 개선시킬 수 있음을 보인다. 그리고 해당 형태로 장면 추출도 가능함을 보이며 장면 추출의 추가 연구 가능성을 시사한다. VCL은 매개변수가 고정된 CLIP에 학습가능한 프롬프트와 MLP로 하이라이트 검출과 장면 추출을 진행한다. 총 2,141개의 학습가능한 매개변수를 사용하여 하이라이트 검출의 HIT@1(>=Very Good) 성능을 기존 CLIP보다 2.71% 개선된 성능과 최소한의 장면 추출 성능을 보인다.

• 대한설비공학회지:설비저널
• /
• 제20권2호
• /
• pp.89-91
• /
• 1991

• 전자통신동향분석
• /
• 제38권6호
• /
• pp.1-11
• /
• 2023

Large language models seem promising for handling reasoning problems, but their underlying solving mechanisms remain unclear. Large language models will establish a new paradigm in artificial intelligence and the society as a whole. However, a major challenge of large language models is the massive resources required for training and operation. To address this issue, researchers are actively exploring compact large language models that retain the capabilities of large language models while notably reducing the model size. These research efforts are mainly focused on improving pretraining, instruction tuning, and alignment. On the other hand, chain-of-thought prompting is a technique aimed at enhancing the reasoning ability of large language models. It provides an answer through a series of intermediate reasoning steps when given a problem. By guiding the model through a multistep problem-solving process, chain-of-thought prompting may improve the model reasoning skills. Mathematical reasoning, which is a fundamental aspect of human intelligence, has played a crucial role in advancing large language models toward human-level performance. As a result, mathematical reasoning is being widely explored in the context of large language models. This type of research extends to various domains such as geometry problem solving, tabular mathematical reasoning, visual question answering, and other areas.

• 대한한의학회지
• /
• 제45권1호
• /
• pp.203-214
• /
• 2024

Objectives: In the context of competency-based education emphasized in Korean Medicine, this study aimed to develop a pilot version of a CPX (Clinical Performance Examination) Practicing Chatbot utilizing large language models with prompt engineering. Methods: A standardized patient scenario was acquired from the National Institute of Korean Medicine and transformed into text format. Prompt engineering was then conducted using role prompting and few-shot prompting techniques. The GPT-4 API was employed, and a web application was created using the gradio package. An internal evaluation criterion was established for the quantitative assessment of the chatbot's performance. Results: The chatbot was implemented and evaluated based on the internal evaluation criterion. It demonstrated relatively high correctness and compliance. However, there is a need for improvement in confidentiality and naturalness. Conclusions: This study successfully piloted the CPX Practicing Chatbot, revealing the potential for developing educational models using AI technology in the field of Korean Medicine. Additionally, it identified limitations and provided insights for future developmental directions.

• 영어어문교육
• /
• 제11권4호
• /
• pp.121-137
• /
• 2005

The present study aims to examine how the roles of teacher and learners affect the repair patterns of both teacher's and learner's utterances in English as a second language (ESL) classroom discourse. The study analyzed beginning ESL classroom discourse and found that the structure of repair seems to be greatly influenced by the roles of participants in a second language classroom. The teacher's repair work was mainly characterized by self-repair. In contrast, learners' repair sequences were predominantly characterized by other-repair. More specifically, self-initiation by the learner of the trouble source was cooperatively completed by the teacher and the other learners. Other-initiated and other-completed repair was the most prevalent form in the current classroom data, which was carried out by the teacher in both modulated and unmodulated manners. When the trouble sources were mostly concerned with the learners' problems with linguistic competence and information presented in the textbook, other-repair took place in a modulated manner (i.e., recasting and prompting). On the other hand, when dealing with learners' errors with factual knowledge, other-repair was conducted in an unmodulated way (i.e., 'no' plus correction).

• 한국콘크리트학회:학술대회논문집
• /
• 한국콘크리트학회 1994년도 가을 학술발표회 논문집
• /
• pp.194-199
• /
• 1994

Frefab Construction known for durable construction skill prompting high productivity in developed country is not yet settled in Korea. This situation of prefab construction results from lack of skill, specialists and quality control. In introducing skill, all equipments are thoughtlessly imported without inside eudeavor for development. Regardless of production of goods, basic study for production of goods, construction and structure is not abailable. The object of this study is curing method in the production process of PC concrete product. From change of curing temperature and curing period which would be factors of product quality in PC concrete production, and research of optimized steam curing condition from relations between curing condition and strength development, basic data of concrete steam curing method will be presented.