Design an eval harness for incident post-mortems using BLEU/ROUGE that tracks inter-judge agreement across prompt versions on Claude 4.5 Sonnet.
Design an eval harness for incident post-mortems using regex match checks that tracks cost-per-correct-answer across prompt versions on Claude Haiku 4.
Design an eval harness for incident post-mortems using DeepEval metrics that tracks cost-per-correct-answer across prompt versions on Gemini 2.0 Flash.
Design an eval harness for incident post-mortems using semantic similarity that tracks token cost across prompt versions on DeepSeek-R1.
Design an eval harness for incident post-mortems using BERTScore that tracks token cost across prompt versions on Llama 3.1 405B.
Design an eval harness for incident post-mortems using promptfoo assertions that tracks token cost across prompt versions on Mistral Small 3.
Design an eval harness for incident post-mortems using semantic similarity that tracks p95 latency across prompt versions on o1-mini.
Design an eval harness for incident post-mortems using BERTScore that tracks p95 latency across prompt versions on o3-mini.
Design an eval harness for incident post-mortems using promptfoo assertions that tracks accuracy across prompt versions on Command R+.
Design an eval harness for incident post-mortems using human pairwise comparison that tracks accuracy across prompt versions on GPT-4.1.
Design an eval harness for incident post-mortems using factuality with retrieval that tracks F1 score across prompt versions on Claude 3.5 Sonnet.
Design an eval harness for incident post-mortems using embedding distance that tracks F1 score across prompt versions on Claude 4.5 Sonnet.