discuss embeddings leadboard, maybe in the context of:
May 22, 2025
NCI has issued a Request for Information (RFI) to gather feedback on key artificial intelligence (AI) benchmarks and related data sets. This input will play a key role in driving progress in the development, evaluation, and validation of AI technologies for cancer research and care.
Key questions to consider include the following:
What specific cases or tasks in cancer research and care could benefit from high-quality AI benchmarks?
What are the most important features of these benchmarks (considering quality, usefulness, and accessibility)?
Can researchers use or adapt any existing data sets to create benchmarks in cancer research and care?
What are the main obstacles to creating or using AI benchmarks in the cancer research field?
Your ideas will help ensure these benchmarks meet the real-world needs of both the cancer and AI communities.
Submissions are due no later than June 30, 2025, at 11:59 p.m. ET.
ranked embeddings as AI( data )
...
...
the simplest I could imagine. f(embed) is a data driven toolset that evolves over time.
... pls add yours
in-browser usage of the embeddings is a major motivation
not clear what we can/should use as storage backend. Ideas / suggestions ...
One idea could be to have a repository per row ...
Demo deployment with TCGA Pathology reports
...
Next Week: https://arxiv.org/pdf/2310.04475 ?
DEMYSTIFYING EMBEDDING SPACES USING LARGE LANGUAGE MODELS