Independent AI Model Evaluator (Self-Directed Practice) — ongoing daily AI response evaluation and RLHF-style ranking
TextProvided ongoing daily evaluation of large language model (LLM) responses using RLHF-style quality criteria and preference ranking. Assessed outputs for accuracy, tone, helpfulness, fluency, coherence, instruction-following, and safety by running systematic prompt tests across multiple models. Documented strengths and recurring failure patterns to support iterative improvement and red-flag detection of hallucinations and unsafe content. • Ranked competing model outputs based on preference judgments consistent with RLHF feedback methodologies • Identified hallucinations, factual errors, logical inconsistencies, and unsafe content in AI responses • Tested behavior across varied prompt types including edge cases and adversarial inputs • Performed native-English checks for naturalness, clarity, and overall response quality