According to arxiv.org, researchers have published several frameworks addressing different applications of large language models (LLMs) in specialized domains.
A paper published July 2, 2026 proposes an automated framework for checking floor plan compliance in residential buildings. According to the abstract, the system addresses Australian policy reforms including SEPP65, BADS, and SPP7.3, which require evaluating health-related features like daylight access, natural ventilation, privacy, and space efficiency. The framework uses an LLM within a “Rule Engine” to convert textual building codes into executable rules, while a “Data Extraction Engine” segments floor plan images into structured building graphs for evaluation.
Another paper from the same date introduces an LLM-based framework for intent-driven network topology design. According to arxiv.org, the system generates network topologies from natural language requirements through “a constraint-driven pipeline combining hierarchical modeling and systematic validation.” The researchers evaluated the framework across four realistic network scenarios and assessed structural correctness using node and edge F1-scores.
A third paper presents SenseWalk, an interactive system for simulating semantic trajectories using LLM-powered agents. According to the abstract, the system combines LLMs with the social force model “to balance physical plausibility and semantic coherence.”
Finally, a paper on NeuroFilter proposes activation-based guardrails for privacy in LLM agents. According to arxiv.org, these filters use activation probing and are designed to be “computationally efficient” for both single-turn and multi-turn conversational settings.