Senior Research Scientist, Reward Models

As a Senior Research Scientist, you will lead research on novel reward model architectures and training approaches for RLHF. You'll develop and evaluate LLM-based grading and evaluation methods, including rubric-driven approaches. Further, you will research techniques to detect, characterize, and mitigate reward hacking and specification gaming.

Responsibilities Include:

-Designing experiments to understand reward model generalization, robustness, and failure modes

-Collaborating with the Finetuning team to translate research insights into improvements for production training pipelines

-Contributing to research publications, blog posts, and internal documentation