Sign in to confirm you’re not a bot
This helps protect our community. Learn more

Improving Large Language Model by Systematically Improving its Data

Microsoft Reactor
115K subscribers
<__slot-el>
<__slot-el>
463 views Streamed 1 year ago
Labeled data powers AI/ML in the enterprise, but real-world datasets have been found to contain between 7-50% annotation errors. Imperfectly labelled text data hampers ML models’ training (and evaluation) across tasks like intent recognition, entity recognition, and sequence generation. Although pretrained LLMs are equipped with a lot of world knowledge, their performance is adversely affected by noisy training data (as noted by OpenAI). In this talk, we illustrate data-centric techniques to mitigate the effect of label noise without changing any code related to model architecture, hyperparameters, or training. These data quality improvement techniques should thus remain applicable even for future advanced LLMs like GPT-10. Resources for this session: Slides: https://docs.google.com/presentation/... Other resources: https://cleanlab.ai/blog/fine-tune-LLM/ https://colab.research.google.com/git... https://docs.cleanlab.ai/stable/index... [eventID:21923] ...more
...more
Improving Large Language Model by Systematically Improving its Data
Labeled data powers AI/ML in the enterprise, but real-world datasets have been found to contain between 7-50% annotation errors. Imperfectly labelled text data hampers ML models’ training (and evaluation) across tasks like intent recognition, entity recognition, and sequence generation. Although pretrained LLMs are equipped with a lot of world knowledge, their performance is adversely affected by noisy training data (as noted by OpenAI). In this talk, we illustrate data-centric techniques to mitigate the effect of label noise without changing any code related to model architecture, hyperparameters, or training. These data quality improvement techniques should thus remain applicable even for future advanced LLMs like GPT-10. Resources for this session: Slides: https://docs.google.com/presentation/... Other resources: https://cleanlab.ai/blog/fine-tune-LLM/ https://colab.research.google.com/git... https://docs.cleanlab.ai/stable/index... [eventID:21923]

Follow along using the transcript.

Microsoft Reactor

115K subscribers
Live chat replay is not available for this video.