Interactive Planning-Based Hypothesis Generation with LTS+ +

Sohrabi, Shirin; Udrea, Octavian; Riabov, Anton; Hassanzadeh, Oktie

doi:10.1007/978-3-030-38561-3_10

Cited by 2 publications

(2 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In other cases, though, example plans could be very useful, especially if the LLM generates them in a targeted way. Leveraging diverse plans (Sohrabi et al 2016;Katz and Sohrabi 2020) could be particularly useful.…”

Section: Discussion and Future Workmentioning

confidence: 99%

Generalized Planning in PDDL Domains with Pretrained Large Language Models

Silver,

Dan,

Srinivas

et al. 2024

AAAI

View full text Add to dashboard Cite

Recent work has considered whether large language models (LLMs) can function as planners: given a task, generate a plan. We investigate whether LLMs can serve as generalized planners: given a domain and training tasks, generate a program that efficiently produces plans for other tasks in the domain. In particular, we consider PDDL domains and use GPT-4 to synthesize Python programs. We also consider (1) Chain-of-Thought (CoT) summarization, where the LLM is prompted to summarize the domain and propose a strategy in words before synthesizing the program; and (2) automated debugging, where the program is validated with respect to the training tasks, and in case of errors, the LLM is re-prompted with four types of feedback. We evaluate this approach in seven PDDL domains and compare it to four ablations and four baselines. Overall, we find that GPT-4 is a surprisingly powerful generalized planner. We also conclude that automated debugging is very important, that CoT summarization has non-uniform impact, that GPT-4 is far superior to GPT-3.5, and that just two training tasks are often sufficient for strong generalization.

show abstract

Section: Discussion and Future Workmentioning

confidence: 99%

Generalized Planning in PDDL Domains with Pretrained Large Language Models

Silver,

Dan,

Srinivas

et al. 2024

AAAI

View full text Add to dashboard Cite

show abstract

“…Furthermore, there are a number of papers that talk about interactive hypothesis generation. For instance, Sohrabi et al (2020) introduce LTS++, which is an interactive development environment for planning-based hypothesis generation in applications with unreliable observations. According to Yurieva (2019), the hypothesis under consideration here is aimed at understanding ontogenetic cognitive discursive mechanisms involved in narration through experimental research on the interactive component in the oral narratives of preschool children.…”

Section: Theoretical Frameworkmentioning

confidence: 99%

AI Application in Foreign Language Literature: ChatGPT's Impact and Skill Enhancement

Chen,

Gao,

Tang

et al. 2024

EJCEEL

View full text Add to dashboard Cite

This study investigates the prospective of incorporating input hypothesis, output hypothesis, interaction hypothesis and metacognitive theory into the best practices of ChatGPT in foreign literature. The study used an online survey involving 146 Chinese internet users to examine opinions on the use of ChatGPT for foreign literature analysis and tasks. Specific skills enhanced by AI integration, such as critical thinking, nuanced comprehension, and advanced linguistic proficiency in foreign languages are discussed in the research. However, there were concerns about its ethical implications when used for literary purposes because it may lead to biases, wrong information and risks which include plagiarism. During the application of ChatGPT, it was emphasized that comprehensible input is crucial to ensure that AI provided content is both understandable and suitable to learners’ current proficiency level. Moreover, learners are encouraged to actively utilize ChatGPT as feedback as well as correction tool. Furthermore, this research allowed learners to engage in live communication through using ChatGPT as a speaking partner thereby negotiating meaning and getting instant feedback to support interaction hypothesis (IH). Underpinned by metacognitive theory, instructing learners to reflect on their learning process with the help of ChatGPT involves identifying areas of difficulty and seeking specific assistance, for example, learners may ask about linguistic doubts from chat GTP or test their understanding about certain concepts or vocabulary using it. In order to effectively integrate these hypotheses with metacognitive theories, interactive chat GPTs were designed which required learners’ involvement in comprehensible input generation of language, output meaningful interactions and reflection on learning process; reading or listening to foreign language content (input), responding or creating content (output), engaging in conversation (interaction) tasks are assessed for their comprehension and performance levels (metacognition).

show abstract

Interactive Planning-Based Hypothesis Generation with LTS+ +

Cited by 2 publications

References 17 publications

Generalized Planning in PDDL Domains with Pretrained Large Language Models

Generalized Planning in PDDL Domains with Pretrained Large Language Models

AI Application in Foreign Language Literature: ChatGPT's Impact and Skill Enhancement

Contact Info

Product

Resources

About