This paper introduces a novel algorithm for constructing decision trees using large language models (LLMs) in zero-shot manner based on Classification and Regression Trees (CART) principles. Traditional tree induction methods rely heavily labeled data to recursively partition criteria such as information gain or the Gini index. In contrast, we propose method that uses pre-trained knowledge embedded LLMs build without requiring training data. Our approach leverages perform operations...

Written answers to open-ended questions can have a higher long-term effect on learning than multiple-choice questions. However, it is critical that teachers immediately review the answers, and ask redo those are incoherent. This be difficult task time-consuming for teachers. A possible solution automate detection of incoherent answers. One option with Large Language Models (LLM). They powerful discursive ability used explain decisions. In this paper, we analyze responses fourth graders in...

Abstract Arguing and communicating is one of the basic skills mathematics curriculum. Making them in written form facilitates rigorous reasoning. It allows peers to review argumentation, receive feedback from them. Even though they generate additional cognitive efforts calculation process, enhance long-term retention facilitate deeper understanding. However, developing arguing competences elementary school classrooms are a great challenge. requires at least two conditions: all students write...

Arguing and communicating are basic skills in the mathematics curriculum. Making arguments written form facilitates rigorous reasoning. It allows peers to review arguments, receive feedback about them. Even though it requires additional cognitive effort calculation process, enhances long-term retention deeper understanding. However, developing these competencies elementary school classrooms is a great challenge. at least two conditions: all students write immediate feedback. One solution use...

We propose a general method to break down main complex task into set of intermediary easier sub-tasks, which are formulated in natural language as binary questions related the final target task. Our allows for representing each example by vector consisting answers these questions. call this representation Natural Language Learned Features (NLLF). NLLF is generated small transformer model (e.g., BERT) that has been trained Inference (NLI) fashion, using weak labels automatically obtained from...

Predicting long-term student achievement is a critical task for teachers and educational data mining. However, most of the models do not consider two typical situations in real-life classrooms. The first that develop their own questions online formative assessment. Therefore, there are huge number possible questions, each which answered by only few students. Second, assessment often involves open-ended students answer writing. These types highly valuable. analyzing responses automatically...

: Predicting long-term student learning is a critical task for teachers and educational data mining. However, most of the models do not consider two typical situations in real-life classrooms. The first that develop their own questions formative assessment. Therefore, there are huge number possible questions, each which answered by only few students. Second, assessment often involved open-ended students answer writing. These types highly valuable. analyzing responses automatically can be...

Predicting long-term student achievement is a critical task for teachers and educational data mining. However, most of the models do not consider two typical situations in real-life classrooms. The first that develop their own questions online formative assessment. Therefore, there are huge number possible questions, each which answered by only few students. Second, assessment often involves open-ended students answer writing. These types highly valuable. analyzing responses automatically...

Written answers to open-ended questions can have a higher long-term effect on learning than multiple-choice questions. However, it is critical that teachers immediately review the answers, and ask redo those are incoherent. This be difficult task time-consuming for teachers. A possible solution automate detection of incoherent answers. One option with Large Language Models (LLM). In this paper, we analyze responses fourth graders in mathematics using three LLMs: GPT-3, BLOOM, YOU. We used...

We propose a general method to break down main complex task into set of intermediary easier sub-tasks, which are formulated in natural language as binary questions related the final target task. Our allows for representing each example by vector consisting answers these questions. call this representation Natural Language Learned Features (NLLF). NLLF is generated small transformer model (e.g., BERT) that has been trained Inference (NLI) fashion, using weak labels automatically obtained from...