Current browse context:
cs.SE
Change to browse by:
References & Citations
Computer Science > Software Engineering
Title: Do Large Language Models Pay Similar Attention Like Human Programmers When Generating Code?
(Submitted on 2 Jun 2023 (v1), last revised 23 May 2024 (this version, v2))
Abstract: Large Language Models (LLMs) have recently been widely used for code generation. Due to the complexity and opacity of LLMs, little is known about how these models generate code. We made the first attempt to bridge this knowledge gap by investigating whether LLMs attend to the same parts of a task description as human programmers during code generation. An analysis of six LLMs, including GPT-4, on two popular code generation benchmarks revealed a consistent misalignment between LLMs' and programmers' attention. We manually analyzed 211 incorrect code snippets and found five attention patterns that can be used to explain many code generation errors. Finally, a user study showed that model attention computed by a perturbation-based method is often favored by human programmers. Our findings highlight the need for human-aligned LLMs for better interpretability and programmer trust.
Submission history
From: Zhijie Wang [view email][v1] Fri, 2 Jun 2023 00:57:03 GMT (1993kb,D)
[v2] Thu, 23 May 2024 17:27:12 GMT (362kb,D)
Link back to: arXiv, form interface, contact.