sgo - Semantic Gradient Optimization
Optimize entities against evaluator populations using LLMs and counterfactual probes.
Tags
Updated: 2026-09-16Capabilities
Typical Inputs
Typical Outputs
What this skill does
- Download persona dataset
- Build entity document
- Filter persona data
- Sample persona cohort
- Generate cohort using LLM
- Evaluate entity against cohort
- Probe counterfactual changes
- Compare evaluation runs
- Audit evaluator cognitive biases
Inputs
- Entity file path or description
- LLM API key
- Cohort filter parameters
- Changes configuration file
- Run tags
Outputs
- Entity Markdown file
- Cohort JSON file
- Evaluation summary and segment breakdown
- Semantic gradient ranking
- Bias audit Markdown report
Requirements
- SGO repository installation
- Python environment with uv or pip
- LLM API key configured in .env file