Be a part of our community of 1M+ users who develop and demonstrate their skills on CodeSignal
4 courses
51 practices
4 hours
Master practical LLM evaluation by benchmarking models on QA and text generation, using metrics like fuzzy matching, ROUGE, and semantic similarity. Learn to analyze logprobs, perplexity, and model behavior for robust, real-world NLP assessment.
Learn how to benchmark large language models using multiple-choice QA, summarization, and scoring techniques like fuzzy matching, ROUGE, and semantic similarity. Compare GPT models across tasks and dive into internal evaluation with log probabilities and perplexity.
This course explores benchmarking for open-ended generation tasks like summarization. You'll experiment with different prompting styles, compare models like GPT-3.5 and GPT-4, and evaluate results using both fuzzy string similarity and semantic similarity via embeddings.
In this course, you'll explore how to evaluate the fluency and likelihood of LLM outputs using internal scoring signals like log probabilities and perplexity. You'll work with OpenAI's completion models to analyze how models "think" under the hood. This course builds naturally on the first two by focusing on model-internal evaluation instead of external references.
In this course, you’ll experiment with deeper aspects of LLM evaluation: token usage efficiency, temperature sensitivity, model output consistency, and detecting hallucinations. Through lightweight API experiments, you’ll develop intuition for how models behave beyond accuracy scores.
I love that it's personalized. When I'm stuck, I don't have to hope my Google searches come out successful. The AI mentor Cosmo knows exactly what I need.
Faith Yim
Software Engineer
+14
It's an amazing product and exceeded my expectations, helping me prepare for my job interviews. Hands-on learning requires you to actually know what you are doing.
Alex Bush
Full Stack Engineer
+9
I'm really impressed by the AI tutor Cosmo's feedback about my code. It's honestly kind of insane to me that it's so targeted and specific.
Abbey Helterbran
Tech consultant
+8
I tried Leetcode but it was too disorganized. CodeSignal covers all the topics I'm interested in and is way more structured.
Jonathan Miller
Senior Machine Learning Engineer
+12
I'm impressed by the quality and can't stop recommending it. It's also a lot of fun!
