Guardado en:
| Autor principal: | |
|---|---|
| Formato: | Recurso digital |
| Lenguaje: | |
| Publicado: |
Zenodo
2025
|
| Materias: | |
| Acceso en línea: | https://doi.org/10.5281/zenodo.17796407 |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
Tabla de Contenidos:
- <p>Large language models (LLMs) typically respond to prompts such as “act as an expert” or “explain step-by-step,” which modify surface-level behavior but do not alter the underlying structure of reasoning.</p> <p>This study introduces a Time-First generative sequence (Time → Space → Consciousness) as a <strong>structural basis for reasoning</strong>, and examines how LLMs behave when this structure is applied directly. Unlike conventional prompting, this approach provides a <em>generative principle</em> that governs the starting point, direction, and stability of reasoning.</p> <p>Across four different models—ChatGPT, Gemini, Claude, and Perplexity—the following common patterns were observed:</p> <ul> <li> <p>Reasoning converges into a consistent flow: “process → stabilization → evaluation”</p> </li> <li> <p>Higher coherence and reproducibility, with increased depth in structural interpretation</p> </li> <li> <p>Responses shift from surface explanations to causal, structural reinterpretations</p> </li> <li> <p>Significant improvements in originality and extendability compared with baseline prompts</p> </li> </ul> <p>These observations suggest that a Time-First Structural Prompt can influence LLM reasoning at a foundational level. The findings provide preliminary insights toward a new research direction: <strong>Structural Prompting</strong>, which aims to modify the generative principles behind LLM thought processes.</p>