mlsanityData to 7 Oct 2026

Taxonomy of Real Faults in Deep Learning Systems

arXiv 1910.11015 ↗First posted 24 Oct 2019cs.SE cs.AI cs.LGcited by 90

Not on the map. Why?

Abstract (arXiv)

The growing application of deep neural networks in safety-critical domains makes the analysis of faults that occur in such systems of enormous importance. In this paper we introduce a large taxonomy of faults in deep learning (DL) systems. We have manually analysed 1059 artefacts gathered from GitHub commits and issues of projects that use the most popular DL frameworks (TensorFlow, Keras and PyTorch) and from related Stack Overflow posts. Structured interviews with 20 researchers and practitioners describing the problems they have encountered in their experience have enriched our taxonomy with a variety of additional faults that did not emerge from the other two sources. Our final taxonomy was validated with a survey involving an additional set of 21 developers, confirming that almost all fault categories (13/15) were experienced by at least 50% of the survey participants.

Key facts. Cited by 90 papers in the map (data through 7 Oct 2026). First posted on arXiv 24 Oct 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 13; 2022: 15; 2023: 15; 2024: 17; 2025: 12; 2026: 18.

90 papers.

Sort:

By area:

  1. 2024Bugs in Large Language Models Generated Code: An Empirical Studycited by 59
  2. 2021Collaboration Challenges in Building ML-Enabled Systems: Communication, Documentation, Engineering, and Processcited by 56
  3. 2021Mind the Gap! A Study on the Transferability of Virtual vs Physical-world Testing of Autonomous Driving Systemscited by 36
  4. 2026ReliabilityBench: Evaluating LLM Agent Reliability Under Production-Like Stress Conditionscited by 31
  5. 2021DeepHyperion: Exploring the Feature Space of Deep Learning-Based Systems through Illumination Searchcited by 29
  6. 2021Silent Bugs in Deep Learning Frameworks: An Empirical Study of Keras and TensorFlowcited by 25
  7. 2021Characterizing Technical Debt and Antipatterns in AI-Based Systems: A Systematic Mapping Studycited by 21
  8. 2021An Empirical Study on Deployment Faults of Deep Learning Based Mobile Applicationscited by 20
  9. 2021Faults in Deep Reinforcement Learning Programs: A Taxonomy and A Detection Approachcited by 18
  10. 2022When and Why Test Generators for Deep Learning Produce Invalid Inputs: an Empirical Studycited by 18
  11. 2022DeepFD: Automated Fault Diagnosis and Localization for Deep Learning Programscited by 16
  12. 2022Bugs in Machine Learning-based Systems: A Faultload Benchmarkcited by 15
38 more
  1. 2024Are LLMs Correctly Integrated into Software Systems?cited by 13
  2. 2021DeepMetis: Augmenting a Deep Learning Test Set to Increase its Mutation Scorecited by 12
  3. 2021Understanding Performance Problems in Deep Learning Systemscited by 12
  4. 2023Testing of Deep Reinforcement Learning Agents with Surrogate Modelscited by 12
  5. 2023Challenges and Practices of Deep Learning Model Reengineering: A Case Study on Computer Visioncited by 11
  6. 2024Predicting Safety Misbehaviours in Autonomous Driving Systems using Uncertainty Quantificationcited by 11
  7. 2021Design Smells in Deep Learning Programs: An Empirical Studycited by 10
  8. 2023Bug Characterization in Machine Learning-based Systemscited by 10
  9. 2021Automatic Fault Detection for Deep Learning Programs Using Graph Transformationscited by 9
  10. 2022Black-box Safety Analysis and Retraining of DNNs based on Feature Extraction and Clusteringcited by 8
  11. 2022Generating and Detecting True Ambiguity: A Forgotten Danger in DNN Supervision Testingcited by 8
  12. 2022Code Smells for Machine Learning Applicationscited by 7
  13. 2022Quality issues in Machine Learning Software Systemscited by 7
  14. 2022Requirements Engineering for Machine Learning: A Review and Reflectioncited by 7
  15. 2024A Survey on Failure Analysis and Fault Injection in AI Systemscited by 7
  16. 2024Understanding the Fundamental Design Decisions of Retrieval-Augmented Generation Systemscited by 7
  17. 2024Towards Enhancing the Reproducibility of Deep Learning Bugs: An Empirical Studycited by 6
  18. 2026When Agents Fail: A Comprehensive Study of Bugs in LLM Agents with Automated Labelingcited by 6
  19. 2023Mutation-based Fault Localization of Deep Neural Networkscited by 5
  20. 2026Real Faults in Model Context Protocol (MCP) Software: a Comprehensive Taxonomycited by 5
  21. 2023Mutation Testing of Deep Reinforcement Learning Based on Real Faultscited by 4
  22. 2023An investigation of challenges encountered when specifying training data and runtime monitors for safety critical ML applicationscited by 4
  23. 2024Benchmarking Generative AI Models for Deep Learning Test Input Generationcited by 4
  24. 2025"My productivity is boosted, but ..." Demystifying Users' Perception on AI Coding Assistantscited by 4
  25. 2022Exploring Effects of Computational Parameter Changes to Image Recognition Systemscited by 3
  26. 2023Reliability Assurance for Deep Neural Network Architectures Against Numerical Defectscited by 3
  27. 2023DeltaNN: Assessing the Impact of Computational Environment Parameters on the Performance of Image Recognition Modelscited by 3
  28. 2023Semantic-Based Neural Network Repaircited by 3
  29. 2023NeuroCodeBench: a plain C neural network benchmark for software verificationcited by 3
  30. 2024Automated Repair of AI Code with Large Language Models and Formal Verificationcited by 3
  31. 2024Targeted Deep Learning System Boundary Testingcited by 3
  32. 2024Demystifying Issues, Causes and Solutions in LLM Open-Source Projectscited by 3
  33. 2025Metamorphic Testing for Pose Estimation Systemscited by 3
  34. 2026Will It Survive? Deciphering the Fate of AI-Generated Code in Open Sourcecited by 3
  35. 2026Agent Skills Can Be Harmful: An Empirical Study of Skill-Induced Failures in LLM Agentscited by 3
  36. 2022Challenges in Migrating Imperative Deep Learning Programs to Graph Execution: An Empirical Studycited by 2
  37. 2022Testing Feedforward Neural Networks Training Programscited by 2
  38. 2022Towards Top-Down Automated Development in Limited Scopes: A Neuro-Symbolic Framework from Expressibles to Executablescited by 2

50 most cited of 90 All 90 (JSON)