Semantic Prompting Framework Enhances Spatial-Textual Generation with LLMs
A new framework known as Semantic Prompting has been developed by researchers to facilitate spatial refinement, allowing for the gradual creation of narratives from interactive layouts utilizing large language models (LLMs). This framework tackles three significant issues in current spatial-textual generation: misalignment between interaction and revision, discrepancies in intent between humans and LLMs, and insufficient customization options. Semantic Prompting identifies semantic interactions, deduces refinement intentions, and executes specific positional adjustments. This framework was integrated into a system named S-PRISM. Evaluation results showed that S-PRISM significantly improved the accuracy of interaction-revision refinement. A user study involving 14 participants demonstrated their use of the system for spatial semantic interaction. The findings were published on arXiv under ID 2604.19971.
Key facts
- Semantic Prompting is a framework for spatial refinement using LLMs.
- It addresses interaction-revision misalignment, human-LLM intent misalignment, and lack of granular customization.
- The framework perceives semantic interactions, reasons about refinement intent, and performs targeted positional revisions.
- S-PRISM is the implementation of the Semantic Prompting framework.
- Empirical evaluation showed enhanced precision of interaction-revision refinement.
- A user study with 14 participants was conducted.
- The paper is available on arXiv with ID 2604.19971.
- The work focuses on incremental narrative generation from spatial layouts.
Entities
Institutions
- arXiv