Machine learning pipelines synthesis with large language models
Keywords: Data analysis pipelines, Large Language models, Machine learning pipelines synthesis, Natural language descriptions
Abstract: In the realm of machine learning, the ability to seamlessly translate natural language descriptions into compilable code is a longstanding challenge. This paper presents a novel framework that addresses this challenge by introducing a pipeline capable of iteratively transforming natural language task descriptions into code through high-level machine learning instructions. Central to this framework is the fine-tuning of the LLama model, enabling it to rank different solutions for various problems and select an appropriate fit for a given task. The paper covers the fine-tuning process and provides insights into the general process of transforming natural language descriptions into code. Our approach marks a significant step towards automating code generation, bridging the gap between task descriptions and executable code, and holds promise for advancing machine learning applications across diverse domains. We showcase the effectiveness of our framework through experimental evaluations and discuss its potential applications in various domains, highlighting its implications for advancing the field of machine learning.
Primary Area: representation learning for computer vision, audio, language, and other modalities
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 8255
Loading