Fine Tune Claude 3 Haiku
- Source
- anthropic.com
- Author
- Anthropic News
- Date
Context
Anthropic's article on Claude 3 Haiku in Amazon Bedrock describes letting customers customize the model with their own examples. Per the announcement, a customer prepares prompt-completion pairs, meaning prompts alongside the ideal outputs they want, and a fine-tuning API builds a custom version that can be tested and refined in the Bedrock console or API. This gives teams a way to teach recurring task requirements through examples.
Anthropic's reported example is moderating online forum comments, including identifying insults, threats and explicit content: fine-tuning improved classification accuracy from 81.5% to 99.6% while reducing per query by 85%. That is the company's evidence for one task. The stored text does not state the baseline for the comparison, and other workflows would need their own examples and .
- fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
- eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
Checking sign-in…
Loading comments…