Programmers Are Poor and Overconfident Judges of LLM-Generated Assertions

• URL: http://arxiv.org/abs/2607.08885v1
• Date: Thu, 09 Jul 2026 19:22:35 GMT
• Title: Programmers Are Poor and Overconfident Judges of LLM-Generated Assertions
• Abstract summary: We conduct a controlled experiment with 86 Python programmers to examine developers' ability to assess the correctness and completeness of generated assertions.<n>We find that, contrary to common assumptions, AI assistance may not improve the reliability of code comprehension and review.
• Score: 12.758916403043207
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Code comprehension and code review are already critically important software engineering tasks, and the rising use of AI code generation tools is only increasing that importance. Generative AI has the possibility of supporting these activities, for example by augmenting code with assertions and natural-language explanations describing code behavior. However, little is known about how effective such support may be. We conduct a controlled experiment with 86 Python programmers and a follow-up think-aloud study to examine developers' ability to assess the correctness and completeness of generated assertions of varying quality, and to investigate how natural-language explanations influence these assessments. While programmers can somewhat accurately judge correct assertions (74% accuracy), they perform poorly when shown incorrect assertions (49% accuracy), despite reporting similar levels of confidence in both judgments. This difference in judgment accuracy is statistically significant (p < 0.001): the odds of a developer accurately judging a correct assertion was nearly three times higher than the odds of accurately judging an incorrect assertion (OR = 2.94). Surprisingly, natural-language explanations of assertions provided no overall benefit. Furthermore, low-quality explanations could impair specification assessment accuracy (p = 0.037, OR = 0.58) while simultaneously increasing developer confidence (p = 0.005, 3.99/5 vs. 4.25/5). Our findings suggest that, contrary to common assumptions, AI assistance may not improve the reliability of code comprehension and review. More broadly, our findings highlight the importance of helping developers evaluate machine-generated reliability artifacts, in addition to generating them.

Related papers

• Neuro-Formal Verification: Agentic Language-Agnostic Formal Program Reasoning [7.228124845671868]
  Neuro-formal verification (NFV) harnesses that automation for developers of mainstream programming languages.<n>An AI coding agent translates, an established verifier decides, and a question posed in a mainstream language is answered push-button, at empirical accuracy rather than soundness, with a machine-checked proof.<n>Results on a dataset of correct and incorrect Python problems are encouraging compared to an llm-as-judge baseline.
  arXiv  Detail & Related papers  (2026-08-21T18:00:02Z)
• Beyond the Traceback: Using LLMs for Adaptive Explanations of Programming Errors [52.619793817715326]
  We present a multi-stage crowdsourced study evaluating skill-targeted, LLM-generated Python error messages.<n>We show that while LLM-rewritten messages significantly improved subjective evaluations, these perceived gains did not translate into statistically significant improvements in objective debug performance.<n>This highlights a critical human-AI complementarity gap: explanations that feel better to users do not necessarily make them more effective debuggers.
  arXiv  Detail & Related papers  (2026-08-21T09:14:28Z)
• Correct Answers from Sound Reasoning: Verifiable Process Supervision for Language Models [94.68358825189738]
  We propose a post-training framework for verifiable domains that jointly optimize prediction accuracy and reasoning quality.<n>We evaluate VPS on chess, a controlled testbed where reasoning steps can be deterministically verified against engine signals.<n>VPS preserves accuracy while significantly improving reasoning quality, reducing win-rate error by up to 30% and restoring consistency to near saturation.
  arXiv  Detail & Related papers  (2026-04-03T15:19:46Z)
• Do What? Teaching Vision-Language-Action Models to Reject the Impossible [53.40183895299108]
  Vision-Language-Action (VLA) models have demonstrated strong performance on a range of robotic tasks.<n>We propose Instruct-Verify-and-Act (IVA), a framework that detects when an instruction cannot be executed due to a false premise.<n>Our experiments show that IVA improves false premise detection accuracy by 97.56% over baselines.
  arXiv  Detail & Related papers  (2025-08-22T10:54:33Z)
• Do AI models help produce verified bug fixes? [62.985237003585674]
  Large Language Models are used to produce corrections to software bugs.<n>This paper investigates how programmers use Large Language Models to complement their own skills.<n>The results are a first step towards a proper role for AI and LLMs in providing guaranteed-correct fixes to program bugs.
  arXiv  Detail & Related papers  (2025-07-21T17:30:16Z)
• Understanding Code Understandability Improvements in Code Reviews [79.16476505761582]
  We analyzed 2,401 code review comments from Java open-source projects on GitHub. 83.9% of suggestions for improvement were accepted and integrated, with fewer than 1% later reverted.
  arXiv  Detail & Related papers  (2024-10-29T12:21:23Z)
• AI-enhanced Auto-correction of Programming Exercises: How Effective is GPT-3.5? [0.0]
  This paper investigates the potential of AI in providing personalized code correction and generating feedback. GPT-3.5 exhibited weaknesses in its evaluation, including localization of errors that were not the actual errors, or even hallucinated errors.
  arXiv  Detail & Related papers  (2023-10-24T10:35:36Z)
• When Good and Reproducible Results are a Giant with Feet of Clay: The Importance of Software Quality in NLP [23.30735117217225]
  We present a case study in which we identify and fix three bugs in widely used implementations of the state-of-the-art Conformer architecture. We propose a Code-quality Checklist and release pangoliNN, a library dedicated to testing neural models.
  arXiv  Detail & Related papers  (2023-03-28T17:28:52Z)
• Generation Probabilities Are Not Enough: Uncertainty Highlighting in AI Code Completions [54.55334589363247]
  We study whether conveying information about uncertainty enables programmers to more quickly and accurately produce code. We find that highlighting tokens with the highest predicted likelihood of being edited leads to faster task completion and more targeted edits.
  arXiv  Detail & Related papers  (2023-02-14T18:43:34Z)
• Invalidator: Automated Patch Correctness Assessment via Semantic and Syntactic Reasoning [6.269370220586248]
  In this paper, we propose a novel technique to automatically assess the correctness of APR-generated patches via semantic and syntactic reasoning. We have conducted experiments on a dataset of 885 patches generated on real-world programs in Defects4J. Experiment results show that INVALIDATOR correctly classified 79% overfitting patches, accounting for 23% more overfitting patches being detected by the best baseline.
  arXiv  Detail & Related papers  (2023-01-03T14:16:32Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.