Narrative Landscape: Mapping Narrative Dispositions Across LLMs
For researchers studying LLM behavior, this provides a method to uncover hidden qualitative differences in model outputs that scalar metrics miss.
This paper proposes a quantitative framework to profile LLM dispositions as stable, model-specific regularities, using a narrative constraint-selection task across six frontier models. Results reveal a rigidity-exploration spectrum and show that instruction types shift selection geometry even when scalar metrics are similar.
This study proposes a quantitative framework for profiling LLM dispositions as stable, model-specific regularities in output under repeated, controlled elicitation. Using a structured narrative constraint-selection task administered across six frontier models and three instruction types, we operationalize disposition through two dimensions: "consistency", measured as cross-replication selection overlap via Jaccard similarity, and "diversity", measured as dispersion across options via the inverse Simpson index. We further introduce Narrative Landscape, a PCA-based visualization that maps each model's selection profile into a shared space for direct comparison. Results reveal a clear rigidity-exploration spectrum across model families and show that instruction types shift the geometry of selection spaces even when scalar metrics appear similar, indicating that comparable scores can mask qualitatively distinct selection topologies.