advanced-evaluation

LLM output quality is evaluated consistently using a structured pipeline that supports direct scoring, pairwise comparison, and rubric generation to ensure reliable assessments across text outputs.

LLM-as-Judge is an evaluation method where an LLM assesses text quality against defined criteria. It works by applying a structured metrics framework to score generated outputs reliably.

Bias in LLM evaluation pipelines is mitigated by actively addressing position, length, self-enhancement, verbosity, and authority biases directly within the assessment framework.

Yes, rubric generation can evaluate LLM code reviews by creating standardized criteria for code quality. This provides developers with structured tools to evaluate code consistently and reliably.

You need the advanced-evaluation skill for core LLM evaluation functionality. The context-fundamentals and tool-design skills are possibly required for full functionality depending on your specific pipeline requirements.

Direct scoring for LLM evaluation is susceptible to self-enhancement and verbosity biases. These limitations are addressed by pairing direct scoring with pairwise comparison and bias mitigation strategies.