LLMCoherence Var 100: Validated Parametric Ladders for Preference Coherence

A dataset from the MINT Lab study of LLM preference coherence over parametric outcome ladders. It contains 100 validated ladders across 12 value categories, each structured as a 7-tier scale. The dataset was authored by MINTLABJHUANU and last updated on June 17, 2026.

Use Cases

• Benchmarking LLM preference coherence based on the 100 validated parametric ladders.
• Comparing decision-making patterns across 16 subject models, including GPT-5.4, Opus-4.6, and GLM-4.5 variants.
• Studying how LLMs handle value trade-offs within the 12 described value categories.
• Analyzing the structure and validation of the 7-tier outcome ladders used in the study.

Strengths

• Contains 100 validated parametric ladders, providing a structured testbed.
• Evaluates 16 distinct subject models, enabling broad comparison.
• Ladders are organized across 12 value categories, suggesting thematic diversity.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update was June 17, 2026.

Provenance

Source

MINTLABJHUANU

Collection Method

Created to accompany a study on LLM preference coherence.

Freshness

Last updated 2026-06-17 17:24:59