A
argbe.tech - news
1min read

SageMaker AI Adds a Dynamic Rubric LLM Judge Powered by Amazon Nova

Amazon SageMaker AI introduced a rubric-based LLM judge powered by Amazon Nova that generates prompt-specific evaluation criteria. The workflow helps teams compare multiple model outputs with consistent, programmatic scoring.

AWS expanded Amazon SageMaker AI with an Amazon Nova–powered LLM judge that scores generations using dynamically created rubrics.

  • The judge generates evaluation criteria per prompt, avoiding a one-size-fits-all, static rule set.
  • It can assess outputs from both AI models and human sources using the same rubric-driven approach.
  • Teams can run evaluation workflows and compare multiple LLM responses via SageMaker training jobs.
  • The setup supports systematic measurement of relative performance across different generative AI products.