Research Scientist, Frontier, Zurich
Is deze vacature iets voor u?
Mijn cv maken Maak uw cv en ontdek uw matchpercentage met deze functie — en met alle andere.
De functie
Design and validate novel post-training pipelines (SFT, RLHF, RLAIF) for frontier-class models where no teacher model exists.
Lead research into next-generation Reward Models, exploring architectures, reducing reward hacking, and improving data signal-to-noise ratios.
Develop methods to enhance internal reasoning and Chain-of-Thought capabilities, emphasizing correctness and multi-step problem solving.
Revamp RL paradigms and prompts to maximize performance while maintaining alignment and safety.
Create robust mechanisms to transform user signals into training data, building a flywheel that scales without regression or bias.
Collaborate across teams to apply these recipes to various sizes and modalities, including audio and multimodal tasks.
Lead research into next-generation Reward Models, exploring architectures, reducing reward hacking, and improving data signal-to-noise ratios.
Develop methods to enhance internal reasoning and Chain-of-Thought capabilities, emphasizing correctness and multi-step problem solving.
Revamp RL paradigms and prompts to maximize performance while maintaining alignment and safety.
Create robust mechanisms to transform user signals into training data, building a flywheel that scales without regression or bias.
Collaborate across teams to apply these recipes to various sizes and modalities, including audio and multimodal tasks.
Bekijk de volledige vacature
Taken, profiel, vaardigheden en voordelen — maak gratis een account aan.
Al een account? Inloggen
Vergelijkbare vacatures
Andere functies die kunnen passen.