Tsukada Lab. Computer Network Group

LLM and Blockchain-Empowered Vibe Manufacturing: Framework, Operation and Incentive Design

Bo Qian, Jiacheng Chen, Hanlin Wu, Zeyu Sun, Jingbo Liu, Manabu Tsukada, Yusheng Ji

Advanced Engineering Informatics, Nov 13, 2026

Summary in three lines

  • Vibe manufacturing uses natural-language requests to coordinate multi-factory production.
  • An LLM parses the request into a process graph, which deterministic functions turn into smart contracts.
  • Experiments measure parsing accuracy and deployment cost, and the paper derives payment contracts under different information conditions.

Abstract

Large language models (LLMs) have recently emerged as a promising enabler for smart manufacturing by facilitating natural-language-based interaction, knowledge integration, and decision support. With the growing trend of LLM-assisted everything, the process of making products can be reshaped as vibe manufacturing, where idea-to-product transformation is simplified into just describing the demand to LLM tools. To achieve this vision, in this paper, we propose a vibe manufacturing framework operated on blockchain, which provides trust and traceability for the members in the ephemeral multi-factory production (MFP) network. Specifically, we design formal prompts that constrain the LLM to act as a semantic parser converting the leader's intent into a structured workflow representation, from which executable smart contracts are synthesized by deterministic code-generation functions. To further address the incentive challenges in vibe manufacturing, we develop a contract-theoretic approach that explicitly derives a time-dependent payment function for each smart contract under three different levels of information, namely symmetric information, partially asymmetric information (hidden action), and asymmetric information (hidden action and ability). We prove the global optimality with closed-form solutions for the designed contracts with individual rationality and incentive compatibility guarantee. A proof-of-concept prototype demonstrates the feasibility of the framework, and an evaluation of the LLM-based workflow extraction, measurements of the on-chain cost, and extensive simulations of the contract design demonstrate its effectiveness.

How to cite

B. Qian, J. Chen, H. Wu, Z. Sun, J. Liu, M. Tsukada, and Y. Ji, “LLM and Blockchain-Empowered Vibe Manufacturing: Framework, Operation and Incentive Design,” Advanced Engineering Informatics, 2026.

BibTeX

@article{Qian2026,
  title         = {LLM and Blockchain-Empowered Vibe Manufacturing: Framework, Operation and Incentive Design},
  author        = {Bo Qian and Jiacheng Chen and Hanlin Wu and Zeyu Sun and Jingbo Liu and Manabu Tsukada and Yusheng Ji},
  year          = {2026},
  date          = {2026-11-13},
  urldate       = {2026-11-13},
  journal       = {Advanced Engineering Informatics},
  abstract      = {Large language models (LLMs) have recently emerged as a promising enabler for smart manufacturing by facilitating natural-language-based interaction, knowledge integration, and decision support. With the growing trend of LLM-assisted everything, the process of making products can be reshaped as vibe manufacturing, where idea-to-product transformation is simplified into just describing the demand to LLM tools. To achieve this vision, in this paper, we propose a vibe manufacturing framework operated on blockchain, which provides trust and traceability for the members in the ephemeral multi-factory production (MFP) network. Specifically, we design formal prompts that constrain the LLM to act as a semantic parser converting the leader's intent into a structured workflow representation, from which executable smart contracts are synthesized by deterministic code-generation functions. To further address the incentive challenges in vibe manufacturing, we develop a contract-theoretic approach that explicitly derives a time-dependent payment function for each smart contract under three different levels of information, namely symmetric information, partially asymmetric information (hidden action), and asymmetric information (hidden action and ability). We prove the global optimality with closed-form solutions for the designed contracts with individual rationality and incentive compatibility guarantee. A proof-of-concept prototype demonstrates the feasibility of the framework, and an evaluation of the LLM-based workflow extraction, measurements of the on-chain cost, and extensive simulations of the contract design demonstrate its effectiveness.},
  keywords      = {Other},
  pubstate      = {published},
  tppubtype     = {article}
}