Summary of Is Programming by Example Solved By Llms?, By Wen-ding Li et al.
Is Programming by Example solved by LLMs?
by Wen-Ding Li, Kevin Ellis
First submitted to arxiv on: 12 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Machine Learning (cs.LG); Programming Languages (cs.PL); Software Engineering (cs.SE)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
Summary difficulty | Written by | Summary |
---|---|---|
High | Paper authors | High Difficulty Summary Read the original abstract here |
Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This research paper investigates the capabilities of Large Language Models (LLMs) in Programming-by-Examples (PBE), a type of few-shot inductive inference. The authors experiment with classic domains like lists and strings, as well as an uncommon graphics programming domain. They find that while LLMs are not effective at PBE without fine-tuning, they can achieve higher performance when the test problems are in-distribution. The study analyzes what causes these models to succeed or fail and takes steps toward understanding out-of-distribution generalization. The results suggest that LLMs make significant progress in solving typical PBE tasks, increasing the flexibility and applicability of PBE systems. |
Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine trying to teach a computer how to solve problems by showing it examples. This is called Programming-by-Examples (PBE). Researchers are interested in whether computers can learn from these examples as well as humans do. They tested powerful computer models, called Large Language Models (LLMs), on different types of problems. Surprisingly, the LLMs didn’t do well at first, but got much better when they were fine-tuned for specific tasks. The study helps us understand why this is the case and how we can improve these computers’ abilities to learn from examples. |
Keywords
» Artificial intelligence » Few shot » Fine tuning » Generalization » Inference