An official website of the United States government
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.
more »
« less
This content will become publicly available on April 26, 2026
Towards Understanding the Characteristics of Code Generation Errors Made by Large Language Models
Large Language Models (LLMs) have demonstrated unprecedented capabilities in code generation. However, there remains a limited understanding of code generation errors that LLMs can produce. To bridge the gap, we conducted an in-depth analysis of code generation errors across six representative LLMs on the HumanEval dataset. Specifically, we first employed open coding and thematic analysis to distill a comprehensive taxonomy of code generation errors. We analyzed two dimensions of error characteristics -- semantic characteristics and syntactic characteristics. Our analysis revealed that LLMs often made non-trivial, multi-line code generation errors in various locations and with various root causes. We further analyzed the correlation between these errors and task complexity as well as test pass rate. Our findings highlighted several challenges in locating and fixing code generation errors made by LLMs. In the end, we discussed several future directions to address these challenges.
more »
« less
- Award ID(s):
- 2340408
- PAR ID:
- 10617730
- Publisher / Repository:
- IEEE
- Date Published:
- ISBN:
- 979-8-3315-0569-1
- Page Range / eLocation ID:
- 2587 to 2599
- Format(s):
- Medium: X
- Location:
- Ottawa, ON, Canada
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Recent advances in large language models (LLMs) have transformed software development by automatically generating code from natural language. Yet challenges remain in generating fully correct code that aligns with user intent. Our study reveals that LLMs tend to pay less attention to user prompts as more code tokens are generated. We hypothesize that this attention dilution issue is an important reason for code generation errors. To mitigate this issue, we propose Selective Prompt Anchoring (SPA) to guide code LLMs to pay more attention to user intent when generating code. We evaluate SPA using six base LLMs across six benchmarks. Our results demonstrate that SPA enhances Pass@1 by up to 12.9%, consistently outperforming SOTA methods in all settings. Our code is available at https://github.com/magic-YuanTian/Selective- Prompt-Anchoring.more » « less
-
Large Language Models (LLMs) have demonstrated unprecedented capability in code generation. However, LLM-generated code is still plagued with a wide range of functional errors, especially for complex programming tasks that LLMs have not seen before. Recent studies have shown that developers often struggle with inspecting and fixing incorrect code generated by LLMs, diminishing their productivity and trust in LLM-based code generation. Inspired by the mutual grounding theory in communication, we propose an interactive approach that leverages code comments as a medium for developers and LLMs to establish a shared understanding. Our approach facilitates iterative grounding by interleaving code generation, inline comment generation, and contextualized user feedback through editable comments to align generated code with developer intent. We evaluated our approach on two popular benchmarks and demonstrated that our approach significantly improved multiple state-of-the-art LLMs, e.g., 17.1% pass@1 improvement for code-davinci-002 on HumanEval. Furthermore, we conducted a user study with 12 participants in comparison to two baselines: (1) interacting with GitHub Copilot, and (2) interacting with a multi-step code generation paradigm called Multi-Turn Program Synthesis. Participants completed the given programming tasks 16.7% faster and with 10.5% improvement in task success rate when using our approach. Both results show that interactively refining code comments enables the collaborative establishment of mutual grounding, leading to more accurate code generation and higher developer confidence.more » « less
-
Large language models offer new ways of empowering people to program robot applications-namely, code generation via prompting. However, the code generated by LLMs is susceptible to errors. This work reports a preliminary exploration that empirically characterizes common errors produced by LLMs in robot programming. We categorize these errors into two phases: interpretation and execution. In this work, we focus on errors in execution and observe that they are caused by LLMs being “forgetful” of key information provided in user prompts. Based on this observation, we propose prompt engineering tactics designed to reduce errors in execution. We then demonstrate the effectiveness of these tactics with three language models: ChatGPT, Bard, and LLaMA-2. Finally, we discuss lessons learned from using LLMs in robot programming and call for the benchmarking of LLM-powered end-user development of robot applications.more » « less
-
Large Language Models (LLMs) have made significant strides in various intelligent tasks but still struggle with complex action reasoning tasks that require systematic search. To address this limitation, we introduce a method that bridges the natural language understanding capability of LLMs with the symbolic reasoning capability of action languages---formal languages for reasoning about actions. Our approach, termed {\sf LLM+AL}, leverages the LLM's strengths in semantic parsing and commonsense knowledge generation alongside the action language's expertise in automated reasoning based on encoded knowledge. We compare {\sf LLM+AL} against state-of-the-art LLMs, including {\sc ChatGPT-4}, {\sc Claude 3 Opus}, {\sc Gemini Ultra 1.0}, and {\sc o1-preview}, using benchmarks for complex reasoning about actions. Our findings indicate that while all methods exhibit various errors, {\sf LLM+AL}, with relatively simple human corrections, consistently leads to correct answers, whereas using LLMs alone does not yield improvements even after human intervention. {\sf LLM+AL} also contributes to automated generation of action languages.more » « less
