Cappy: Outperforming and boosting large multi-task language models with a small scorer
How-To How to actually use this
What changed: A new method called Cappy uses a small scorer to improve the performance of large multi-task language models (like T0, FLAN, OPT-IML) that follow instructions.
How to use it:
- Identify a multi-task LLM you are using for instruction-following tasks.
- Apply the Cappy scoring method to evaluate and rank the model's outputs.
- Use the scorer to boost the model's performance on diverse NLP tasks without changing the model itself.
Good for: researchers and developers looking to improve multi-task LLM results with a lightweight add-on.
Posted by Yun Zhu and Lijuan Liu, Software Engineers, Google Research Large language model (LLM) advancements have led to a new paradigm that unifies various natural language processing (NLP) tasks within an instruction-following framework. This paradigm is exemplified by recent multi-task LLMs, such as T0 , FLAN , and OPT-IML . First, multi-task data is gathered with each task following a…
Read original article on Google AI Blog →More in Model Releases
