Aravind Putrevu
Devrel
発見し、つながり、成長する
AI を始める準備はできていますか? Microsoft Reactor は、スタートアップ企業、起業家、開発者が AI テクノロジの上に次のビジネスを構築するのに役立つイベント、トレーニング、コミュニティ リソースを提供します。 ご参加ください。
発見し、つながり、成長する
AI を始める準備はできていますか? Microsoft Reactor は、スタートアップ企業、起業家、開発者が AI テクノロジの上に次のビジネスを構築するのに役立つイベント、トレーニング、コミュニティ リソースを提供します。 ご参加ください。
26 2月, 2024 | 12:00 午後 - 1:00 午後 (UTC) 協定世界時
トピック: データ サイエンスと機械学習
言語: 英語
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.
講演者
ご不明な点がございましたら、お問い合わせください reactor@microsoft.com