About the Summarization Models in Generative AI
The following categories are ideal text sources for summarization:
- News articles
- Blog posts
- Chat transcripts
- Scientific articles
- Meeting notes
- Product reviews
The
cohere.command model supported for the on-demand serving mode is now retired and this model is deprecated for the dedicated serving mode. If you're hosting cohere.command on a dedicated AI cluster, (dedicated serving mode) for summarization, you can continue to use this hosted model replica with the summarization API and in the playground until the cohere.command model retires for the dedicated serving mode. These models, when hosted on a dedicated AI cluster are only available in US Midwest (Chicago). See Retiring the Models for retirement dates and definitions. We recommend that you use the chat models instead which offer the same summarization capabilities, including control over summary length and style.Summarization Model Parameters
When using a hosted summarization model in the playground, you can get a different output by changing the following parameters.
- Length
-
The approximate length of the summary. You can select short, medium, or long. Short summaries are roughly up to two sentences long, medium summaries are between three and five sentences, and long summaries might have six or more sentences. For the Auto value, the model chooses a length based on the input size.
- Format
-
Whether to display the summary in a free-form paragraph or in bullet points. For the Auto value, the model chooses the best format based on the input text.
- Extractiveness
-
How much to reuse the input in the summary. Summaries with high extractiveness tend to use sentences verbatim, and summaries with low extractiveness tend to paraphrase.
- Temperature
-
The level of randomness used to generate the output text.
Tip
To summarize a text, start with the temperature set to 0. If you don't require random results, we recommend a temperature value of 0.2. Use a higher value if, for example, you plan to select various summaries afterward. However, don't use a high temperature for summarization because a high temperature encourages the model to produce creative text, which might also include hallucinations and factually incorrect information. - Additional command
-
Other summarizing options such as style or focus. Write one or more additional commands in a natural language as instructions to the model, for example, "focus on dates", or "write in a conversational style", or "end the resume with END SUMMARY".