Rethinking Software Engineering in the Foundation Model Era: A Curated Catalogue of Challenges in the Development of Trustworthy FMware
Not on the map. Why?
Abstract (arXiv)
Foundation models (FMs), such as Large Language Models (LLMs), have revolutionized software development by enabling new use cases and business models. We refer to software built using FMs as FMware. The unique properties of FMware (e.g., prompts, agents, and the need for orchestration), coupled with the intrinsic limitations of FMs (e.g., hallucination) lead to a completely new set of software engineering challenges. Based on our industrial experience, we identified 10 key SE4FMware challenges that have caused enterprise FMware development to be unproductive, costly, and risky. In this paper, we discuss these challenges in detail and state the path for innovation that we envision. Next, we present FMArts, which is our long-term effort towards creating a cradle-to-grave platform for the engineering of trustworthy FMware. Finally, we (i) show how the unique properties of FMArts enabled us to design and develop a complex FMware for a large customer in a timely manner and (ii) discuss the lessons that we learned in doing so. We hope that the disclosure of the aforementioned challenges and our associated efforts to tackle them will not only raise awareness but also promote deeper and further discussions, knowledge sharing, and innovative solutions across the software engineering discipline.
Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 49 of its 109 references that are in the map. First posted on arXiv 25 Feb 2024.
Builds on
49 of 109 references.
References · 49
- 2022Training language models to follow instructions with human feedbackcited by 15,712
- 2020Retrieval-Augmented Generation for Knowledge-Intensive NLP Taskscited by 9,160
- 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
- 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
- 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
- 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,674
- 2022Survey of Hallucination in Natural Language Generationcited by 2,056
- 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversationcited by 1,906
- 2023ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIscited by 1,880
- 2016Concrete Problems in AI Safetycited by 1,582
- 2020Making Pre-trained Language Models Better Few-shot Learnerscited by 1,255
- 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,107
37 more
- 2022Large Language Models Are Human-Level Prompt Engineerscited by 1,020
- 2023DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelinescited by 803
- 2023Instruction Tuning for Large Language Models: A Surveycited by 617
- 2023Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolutioncited by 521
- 2017Snorkel: Rapid Training Data Creation with Weak Supervisioncited by 370
- 2022RLPrompt: Optimizing Discrete Text Prompts with Reinforcement Learningcited by 361
- 2023The Curse of Recursion: Training on Generated Data Makes Models Forgetcited by 343
- 2023Studying Large Language Model Generalization with Influence Functionscited by 311
- 2021AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Promptscited by 265
- 2023How is ChatGPT's behavior changing over time?cited by 265
- 2023Prompt Cache: Modular Attention Reuse for Low-Latency Inferencecited by 255
- 2023Reasoning or Reciting? Exploring the Capabilities and Limitations of Language Models Through Counterfactual Taskscited by 243
- 2023TaskMatrix.AI: Completing Tasks by Connecting Foundation Models with Millions of APIscited by 184
- 2022Instruction Induction: From Few Examples to Natural Language Task Descriptionscited by 152
- 2022PromptChainer: Chaining Large Language Model Prompts through Visual Programmingcited by 134
- 2023An Empirical Study of the Non-determinism of ChatGPT in Code Generationcited by 134
- 2023Speak, Memory: An Archaeology of Books Known to ChatGPT/GPT-4cited by 123
- 2023Agents: An Open-source Framework for Autonomous Language Agentscited by 116
- 2021Bias in Machine Learning Software: Why? How? What to do?cited by 92
- 2023Use of LLMs for Illicit Purposes: Threats, Prevention Measures, and Vulnerabilitiescited by 76
- 2023The Data Provenance Initiative: A Large Scale Audit of Dataset Licensing & Attribution in AIcited by 73
- 2018Snorkel DryBell: A Case Study in Deploying Weak Supervision at Industrial Scalecited by 52
- 2020Behavioral Use Licensing for Responsible AIcited by 46
- 2023AtMan: Understanding Transformer Predictions Through Memory Efficient Attention Manipulationcited by 30
- 2023Prompt Sapper: A LLM-Empowered Production Tool for Building AI Chainscited by 24
- 2017BoostClean: Automated Error Detection and Repair for Machine Learningcited by 23
- 2023Generative Artificial Intelligence for Software Engineering -- A Research Agendacited by 23
- 2019Towards Standardization of Data Licenses: The Montreal Data Licensecited by 19
- 2023Building Your Own Product Copilot: Challenges, Opportunities, and Needscited by 17
- 2024Multi-line AI-assisted Code Authoringcited by 13
- 2021Active WeaSuL: Improving Weak Supervision with Active Learningcited by 12
- 2023The Importance of Human-Labeled Data in the Era of LLMscited by 8
- 2021Can I use this publicly available dataset to build commercial AI software? -- A Case Study on Publicly Available Image Datasetscited by 6
- 2023Model Reporting for Certifiable AI: A Proposal from Merging EU Regulation into AI Developmentcited by 6
- 2016ActiveClean: Interactive Data Cleaning While Learning Convex Loss Modelscited by 4
- 2022Towards Build Verifiability for Java-based Systemscited by 2
- 2023An investigation of licensing of datasets for machine learning based on the GQM modelcited by 1
Built on by
Citing papers per year: 2024: 3; 2025: 2; 2026: 1.
6 papers.
- 2024Towards AI-Native Software Engineering (SE 3.0): A Vision and a Challenge Roadmapcited by 30
- 2024Towards More Trustworthy and Interpretable LLMs for Code through Syntax-Grounded Explanationscited by 10
- 2024A Taxonomy of Architecture Options for Foundation Model-based Agents: Analysis and Decision Modelcited by 4
- 2025Past, Present, and Future of Bug Tracking in the Generative AI Eracited by 4
- 2025Mapping the Trust Terrain: LLMs in Software Engineering -- Insights and Perspectivescited by 0
- 2026Natural-Language Workflows Are Not Software Yet: Artifact-Driven Compilation for Reliable Agent Executioncited by 0
- 2026Natural-Language Workflows Are Not Software Yet: Artifact-Driven Compilation for Reliable Agent Execution21 Aug 2026
- 2025Past, Present, and Future of Bug Tracking in the Generative AI Era9 Oct 2025
- 2025Mapping the Trust Terrain: LLMs in Software Engineering -- Insights and Perspectives18 Mar 2025
- 2024Towards AI-Native Software Engineering (SE 3.0): A Vision and a Challenge Roadmap8 Oct 2024
- 2024A Taxonomy of Architecture Options for Foundation Model-based Agents: Analysis and Decision Model6 Aug 2024
- 2024Towards More Trustworthy and Interpretable LLMs for Code through Syntax-Grounded Explanations12 Jul 2024
Similar, not cited
Close in wording, no citation either way
- 2024From Cool Demos to Production-Ready FMware: Core Challenges and a Technology Roadmap
- 2026Understanding How Enterprises Adopt the Model Context Protocol for LLM-Driven Software Engineering
- 2025Towards Conversational Development Environments: Using Theory-of-Mind and Multi-Agent Architectures for Requirements Refinement
- 2026An Empirical Study of Model Context Protocol Applications
- 2026From What to How: Bridging User Requirements with Software Development Using Large Language Models
- 2026The Evolution of Tool Use in LLM Agents: From Single-Tool Call to Multi-Tool Orchestration
- 2026Model Context Protocol (MCP) Tool Descriptions Are Smelly! Towards Improving AI Agent Efficiency with Augmented MCP Tool Descriptions
- 2025Challenges in Testing Large Language Model Based Software: A Faceted Taxonomy