Priya Ramanathan

Measure how well a model adheres to specific instructions.

Test model performance at various context lengths.

Analyze token usage patterns to reduce API costs.

Evaluate model safety mechanisms against adversarial inputs.

Measure and analyze model response times for performance optimization.

Screen a model for potential biases across demographic categories.

Identify factual inaccuracies and hallucinations in model outputs without prior examples.

Rapidly evaluate a language model accuracy on a given dataset without examples.