AI Trainer (RLHF Response Evaluation & Ranking) — Outlier AI (Remote)
Executed Reinforcement Learning from Human Feedback (RLHF) tasks by evaluating, comparing, and ranking multiple model responses. Assessed responses using quality, helpfulness, reasoning, factual accuracy, and adherence to project guidelines to support model alignment. Provided detailed preference justifications for multiple RLHF projects to enhance human expectations. • Ranked and preferred among AI-generated responses • Evaluated factuality, reasoning, and helpfulness • Wrote preference justifications for RLHF submissions • Maintained productivity and accuracy across RLHF tasks