AI Trainer
TextRLHF
As an AI Trainer at Outlier AI & Binary Cognition, I evaluated LLM outputs for reasoning quality, factual accuracy, instruction following, and robustness. I provided Reinforcement Learning from Human Feedback (RLHF) preference feedback and created adversarial test cases to stress-test model behavior. My work produced high-quality annotations and assessments used in fine-tuning pipelines.
• Reviewed and rated AI-generated responses to diverse prompts for quality and accuracy.
• Authored adversarial prompts that challenged model capabilities and edge cases.
• Provided RLHF preference judgments to guide LLM tuning and improvement.
• Collaborated with research teams to enhance annotation guidelines and model reliability.
As an AI Trainer at Outlier AI & Binary Cognition, I evaluated LLM outputs for reasoning quality, factual accuracy, instruction following, and robustness. I provided Reinforcement Learning from Human Feedback (RLHF) preference feedback and created adversarial test cases to stress-test model behavior. My work produced high-quality annotations and assessments used in fine-tuning pipelines. • Reviewed and rated AI-generated responses to diverse prompts for quality and accuracy. • Authored adversarial prompts that challenged model capabilities and edge cases. • Provided RLHF preference judgments to guide LLM tuning and improvement. • Collaborated with research teams to enhance annotation guidelines and model reliability.
2022 - Present