Samuel Slezak

2026

1. Yang, X., Slezak, S., Siriwardena, L., Amtsberg, F., & Menges, A. (2026). MAVE: An Augmented Multi-agent LLM System for Interactive Design and Robotic Fabrication. Proceedings of the 2026 Designing Interactive Systems Conference, 76–92. https://doi.org/10.1145/3800645.3813008
   • BibTeX

   BibTeX

   @inproceedings{Yang2026MAVE, abstract = {This research presents MAVE, a framework that combines a multi-agent large language model (LLM) system with an augmented reality (AR) interface to support interactive design and robotic fabrication. While LLM agents are highly flexible, grounding user input in task-relevant digital and physical context remains a key challenge for effective human-machine collaboration. To address this, we introduce five grounding strategies – model, object, spatial, tool, and goal references – that structure how agents interpret and act on user input. We conducted ablation studies to evaluate these strategies, showing that they improve agent execution efficiency and accuracy while reducing the number of interaction turns. We further demonstrate the framework’s adaptability through a design workshop, in which student teams extended a truss design-and-fabrication scenario involving a human and a cobot to prototype diverse interactive workflows. The workshop outcomes highlight the relevance of augmented multi-agent LLM systems for design and fabrication and suggest the framework’s potential to support rapid prototyping of interactive multi-actor workflows.}, address = {New York, NY, USA}, author = {Yang, Xiliu and Slezak, Samuel and Siriwardena, Lasath and Amtsberg, Felix and Menges, Achim}, booktitle = {Proceedings of the 2026 Designing Interactive Systems Conference}, doi = {10.1145/3800645.3813008}, eventdate = {2026-06-13/2026-06-17}, eventtitle = {DIS '26: Designing Interactive Systems Conference}, isbn = {979-8-4007-2563-0}, language = {English}, location = {Singapore, Singapore}, month = {06}, note = {Related to: S. Slezak, X. Yang, L. Siriwardena, F. Amtsberg, and A. Menges. Software, (2026)Related to: X. Yang, S. Slezak, L. Siriwardena, F. Amtsberg, and A. Menges, “MAVE: An Augmented Multi-Agent LLM System for Interactive Design and Robotic Fabrication”, in Proceedings of the 2026 DIS Conference on Designing Interactive Systems, DIS ’26, Singapore: Association for Computing Machinery, 2026. doi: 10.1145/3800645.3813008.}, organization = {SIGCHI ACM Special Interest Group on Computer-Human Interaction}, pages = {76-92}, publisher = {ACM}, series = {DIS '26}, title = {MAVE: An Augmented Multi-agent LLM System for Interactive Design and Robotic Fabrication}, url = {https://doi.org/10.1145/3800645.3813008}, venue = {Singapore, Singapore}, year = 2026 }
2. Slezak, S., Yang, X., Siriwardena, L., Amtsberg, F., & Menges, A. (2026). Data for MAVE: An Augmented Multi-agent LLM System for Interactive Design and Robotic Fabrication. https://doi.org/10.18419/darus-5838
   • BibTeX

   BibTeX

   @dataset{slezak2026augmented, abstract = {This repository is the dataset accompanying the paper: "MAVE: An Augmented Multi-agent LLM System for Interactive Design and Robotic Fabrication". It contains the source code of the MAVE system, the study setup and traces, and analysis scripts needed to reproduce the data in the paper. MAVE includes a hierarchical multi-agent architecture with supervisor, design, and robot agents. The codebase documents how user input, context sensing, geometry manipulation, and robot control are integrated within a common workflow, including the JSON-based model representation, database-supported context logging, and task-specific tool interfaces used by the agents. The repository also includes the test data required to reproduce the evaluation scenarios reported in the paper, particularly the ablation studies examining the contribution of model, object, spatial, tool, and goal grounding strategies. Together, these materials support inspection of the system design, replication of benchmark tasks, and adaptation of the framework to related human-robot collaboration settings in design and fabrication. The repository is intended as a reusable research artefact that promotes transparency, reproducibility, and extensibility. }, affiliation = {Slezak, Samuel/University of Stuttgart, Yang, Xiliu/University of Stuttgart, Siriwardena, Lasath/University of Stuttgart, Amtsberg, Felix/University of Applied Sciences Biberach, Menges, Achim/University of Stuttgart}, author = {Slezak, Samuel and Yang, Xiliu and Siriwardena, Lasath and Amtsberg, Felix and Menges, Achim}, doi = {10.18419/darus-5838}, howpublished = {Software}, note = {Related to: X. Yang, S. Slezak, L. Siriwardena, F. Amtsberg, and A. Menges, “MAVE: An Augmented Multi-Agent LLM System for Interactive Design and Robotic Fabrication”, in Proceedings of the 2026 DIS Conference on Designing Interactive Systems, DIS ’26, Singapore: Association for Computing Machinery, 2026. doi: 10.1145/3800645.3813008}, orcid-numbers = {Slezak, Samuel/https://orcid.org/0009-0003-0303-7730, Yang, Xiliu/https://orcid.org/0000-0002-5835-9898, Siriwardena, Lasath/https://orcid.org/0000-0001-9100-1321, Amtsberg, Felix/https://orcid.org/0000-0002-1105-8297, Menges, Achim/https://orcid.org/0000-0001-9055-4039}, title = {Data for MAVE: An Augmented Multi-agent LLM System for Interactive Design and Robotic Fabrication}, url = {https://doi.org/10.18419/DARUS-5838}, year = 2026 }
3. Slezak, S., Siriwardena, L., Leder, S., & Menges, A. (2026). Replication Data for: LLM-Driven behavior Generation and Negotiation for Collective Robotic Construction. https://doi.org/10.18419/darus-5783
   • BibTeX

   BibTeX

   @dataset{slezak2026replication, abstract = {Supplementary repository for the paper on LLM-driven collective robotic construction (CRC).Agent-based modeling and simulation (ABMS) has been widely employed to study emergent processes in collective robotic construction (CRC), where global architectural structures arise from local agent interactions. While these approaches reveal how complex assemblies can emerge without centralized control, they remain limited when an architectural goal is known but the behaviors required to achieve it are not. Most CRC workflows still depend on hand-crafted heuristics. This paper presents a hybrid CRC workflow that integrates large language models (LLM) into the ABMS behavior design process. The system enables human-AI co-creation of robot behaviors, allowing an LLM agent to generate and negotiate behavioral strategies toward user-defined construction goals under partial observability. The approach is evaluated in simulation, comparing an LLM-heuristic hybrid against a heuristic-only baseline behavior. By embedding language-based reasoning within ABMS, this work expands participation in CRC behavior design and demonstrates a pathway for translating high-level design intent into adaptive, convergent, and goal-oriented multi-agent construction processes. }, affiliation = {Slezak, Samuel/University of Stuttgart, Siriwardena, Lasath/University of Stuttgart, Leder, Samuel/University of Stuttgart, Menges, Achim/University of Stuttgart}, author = {Slezak, Samuel and Siriwardena, Lasath and Leder, Samuel and Menges, Achim}, doi = {10.18419/darus-5783}, howpublished = {Dataset}, orcid-numbers = {Slezak, Samuel/https://orcid.org/0009-0003-0303-7730, Siriwardena, Lasath/https://orcid.org/0000-0001-9100-1321, Leder, Samuel/https://orcid.org/0000-0003-3533-3633, Menges, Achim/https://orcid.org/0000-0001-9055-4039}, title = {Replication Data for: LLM-Driven behavior Generation and Negotiation for Collective Robotic Construction}, url = {https://doi.org/10.18419/DARUS-5783}, year = 2026 }
4. Slezak, S., & Siriwardena, L. (2026). Cassis: An Agentic AI Toolkit for Rhino Grasshopper [DaRUS]. https://doi.org/10.18419/DARUS-5777
   • BibTeX

   BibTeX

   @dataset{https://doi.org/10.18419/darus-5777, abstract = {Cassis consists of the plugin library Cassis and a transport library for Streamable HTTP-based MCP communication for Rhino Grasshopper. Cassis implements the functionality specific to connecting agentic AI systems to Rhino Grasshopper through the Model Context Protocol. The plugin can, in principle, be referenced from any AI coding assistant that supports MCP, including IDE Copilots / Agents, Terminal CLI agents, and proprietary systems. Cassis defines three service interfaces for component manipulation, document management, and UI thread coordination, and the MCP host class. In addition to the service layer, the plugin provides tool categories: script tools for C# and Python components, component tools for adding and configuring Grasshopper objects, connection tools for wiring parameters, panel tools, group tools, canvas and viewport capture tools, document tools, solver control tools, component analytics tools, diagnostics tools, and ABxM interoperability tools for agent-based modelling. The necessary tool, service, and transport classes are derived from the corresponding interfaces and discovered at startup. A Grasshopper component exposes the server on the canvas and lets users select which tools the AI assistant can call. Instructions for the use of the source code are included in the file "README.md". }, author = {Slezak, Samuel and Siriwardena, Lasath}, doi = {10.18419/DARUS-5777}, publisher = {DaRUS}, title = {Cassis: An Agentic AI Toolkit for Rhino Grasshopper}, url = {https://doi.org/10.18419/DARUS-5777}, year = 2026 }
5. Slezak, S., Shevidi, S., Shakeri, Z., Ron, G., Amtsberg, F., Wortmann, T., & Menges, A. (2026). Human–robot collaboration in building disassembly: a multi-agent LLM architecture. Construction Robotics, 10, Article 2. https://doi.org/10.1007/s41693-026-00189-x
   • BibTeX

   BibTeX

   @article{slezak2026humanrobot, abstract = {Building disassembly is critical for circular economy material reuse, yet remains rare due to cost and safety constraints, leading to demolition and material downcycling. Automation could improve both efficiency and safety, but currently available technology does not yet enable full automation. We propose a human–robot collaboration system architecture that uses agentic large language models. We test this approach in building disassembly—an unstructured, safety–critical domain where conventional pre-programmed robotics are inadequate. The agentic architecture combines curated domain knowledge, physics simulation for stability validation, and natural language interfaces, enabling the robot to participate through proactive reasoning rather than follow control commands. We evaluated the architecture through three progressively complex scenarios: collaborative spatial adaptation, collaborative decision-making, and learning. The main contribution is a modular, data-grounded HRC methodology in which specialized LLM agents perform agentic reasoning: the robot assesses situations, retrieves relevant procedural knowledge, validates decisions through simulation, and negotiates solutions with human operators. This proof of concept demonstrates that agentic multi-agent LLM systems can enable adaptive human–robot collaboration under uncertainty, beyond natural language interfaces through integrated domain knowledge, physics validation, and agentic reasoning.}, author = {Slezak, Samuel and Shevidi, Shirin and Shakeri, Zahra and Ron, Gili and Amtsberg, Felix and Wortmann, Thomas and Menges, Achim}, doi = {10.1007/s41693-026-00189-x}, issn = {2509-8780}, journal = {Construction Robotics}, language = {English}, month = {07}, number = 2, pages = 23, publisher = {Springer Science and Business Media LLC}, title = {Human–robot collaboration in building disassembly: a multi-agent LLM architecture}, url = {https://doi.org/10.1007/s41693-026-00189-x}, volume = 10, year = 2026 }
6. Slezak, S., Siriwardena, L., Leder, S., & Menges, A. (2026). LLM-driven behavior generation and negotiation for collective robotic construction. Construction Robotics, 10, Article 2. https://doi.org/10.1007/s41693-026-00216-x
   • BibTeX

   BibTeX

   @article{Slezak2026LLMDriven, author = {Slezak, Samuel and Siriwardena, Lasath and Leder, Samuel and Menges, Achim}, doi = {10.1007/s41693-026-00216-x}, issn = {2509-8780}, journal = {Construction Robotics}, language = {English}, month = {07}, number = 2, pages = 28, publisher = {Springer Science and Business Media LLC}, title = {LLM-driven behavior generation and negotiation for collective robotic construction}, url = {https://doi.org/10.1007/s41693-026-00216-x}, volume = 10, year = 2026 }