|
Work item:
|
H.FDM-AC-Gen
|
|
Subject/title:
|
Assessment criteria for foundation models: General
|
|
Status:
|
[Carried to next study period]
|
|
Approval process:
|
AAP
|
|
Type of work item:
|
Recommendation
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
-
|
|
Liaison:
|
ITU-T SG13, ISO/IEC JTC1 SC 42, IEEE AISC
|
|
Supporting members:
|
-
|
|
Summary:
|
A foundation model is any model that is trained on broad data (generally using self-supervision at scale) that can be adapted (e.g, fine-tuned) to a wide range of downstream tasks. The significance of foundation models can be summarized by two words: emergence and homogenization. Foundation models have capacities in domain of language, vision, robotics, reasoning,search and interaction, indicating that they have the potential to be transformed into various sectors and industries, extending the role of AI in the society. For foundation models, they can be and will be applied widely to enable most of the sectors and industries. As the foundation model's development and its applications are being widely and rapidly used, there are a series of concerns to be studied, discussed and standardized.
This Recommendation specify relevant definitions for foundation models, in the meantime, emergence and homogenization for foundation models will be also described clearly. What's more, it will be described with the framework for assessment for foundation models with specific domains. In the appendix a series of typical UCs will be collected and illustrated.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2023-08-22 16:53:18
|
|
Last update:
2024-09-20 10:30:29
|
|
