Ethan Park

Analyzes disagreements in self-consistency response sets to identify ambiguity sources and resolution strategies.

Generates detailed domain expert personas with appropriate vocabulary, boundaries, and ethical guidelines.

Manages backtracking decisions in tree-of-thought exploration while preserving useful discoveries.

Designs prompts with precise response length controls and content prioritization for overflow handling.

Creates reusable prompt template libraries with parameterized slots and comprehensive documentation.

Detects points where chain-of-thought reasoning could branch into alternative solution paths.

Extracts reusable patterns from few-shot examples to enable generalization to new cases.

Builds zero-shot classification prompts with clear category boundaries and ambiguity handling.

Diagnoses prompt failures through structured reflection comparing expected versus actual outputs.

Maps dependencies between decomposed tasks to optimize scheduling and parallelization opportunities.

Validates that LLM responses maintain consistent persona characteristics and expertise boundaries.

Enhances prompts by inserting explicit chain-of-thought reasoning scaffolds at appropriate locations.