Ethan Park

Analyzes how well model responses follow given instructions with detailed compliance scoring.

Traces and visualizes reasoning paths across self-consistency samples to identify divergence patterns.

Designs merit functions for evaluating tree-of-thought branches with proper weighting and normalization.

Sets ethical boundaries for role personas with appropriate refusal responses and nuanced handling.

Handles numerical precision in outputs with configurable rounding, significant figures, and formatting.

Adapts prompts between different LLM models by mapping capabilities and adjusting instruction styles.

Manages abstraction levels in chain-of-thought reasoning to match audience expertise appropriately.

Generates XML output schemas with proper element hierarchy, namespaces, and validation constraints.

Explores counterfactual alternatives to decisions to extract improvement insights through comparison.

Tunes task decomposition granularity by rebalancing over and under-decomposed subtasks.

Manages long-term memory for role personas including fact extraction, compression, and consistency.

Enforces factual grounding by requiring and validating source references for all claims.