GPT is becoming a Turing machine: Here are some ways to program it

TL;DR

This paper shows that with specific prompting techniques, GPT-3 models can perform iterative programming tasks and execute algorithms, revealing new capabilities and implications for evaluation and education.

Contribution

It introduces methods to trigger iterative and algorithmic behavior in GPT-3 through prompting, demonstrating execution-like performance beyond simple text generation.

Findings

IRSA improves accuracy in program execution tasks

Prompting can induce algorithmic behavior without full task examples

Prompt design critically influences LLM performance

Abstract

We demonstrate that, through appropriate prompting, GPT-3 family of models can be triggered to perform iterative behaviours necessary to execute (rather than just write or recall) programs that involve loops, including several popular algorithms found in computer science curricula or software developer interviews. We trigger execution and description of Iterations by Regimenting Self-Attention (IRSA) in one (or a combination) of three ways: 1) Using strong repetitive structure in an example of an execution path of a target program for one particular input, 2) Prompting with fragments of execution paths, and 3) Explicitly forbidding (skipping) self-attention to parts of the generated text. On a dynamic program execution, IRSA leads to larger accuracy gains than replacing the model with the much more powerful GPT-4. IRSA has promising applications in education, as the prompts and responses resemble student assignments in data structures and algorithms classes. Our findings hold implications for evaluating LLMs, which typically target the in-context learning: We show that prompts that may not even cover one full task example can trigger algorithmic behaviour, allowing solving problems previously thought of as hard for LLMs, such as logical puzzles. Consequently, prompt design plays an even more critical role in LLM performance than previously recognized.