https://microsoft.github.io/LLM2CLIP/
Improved CLIP performance and context window length by using LLMs as a text encoder. This looks very relevant for all projects deploying CLIP for retrieval tasks.
[#]AI #datascience
=> More informations about this toot | View the thread
Here's a #FollowFriday post with some of my recent followings (inspired by a post from @julian ):
@CatherineFlick - Professor of Ethics and Games Technology
@ambv - CPython and black (auto-formatter) developer
@TheodoreKim - Professor of Computer Graphics
@jvinagre - Researcher on AI and recommender systems
@kgajos - Professor of Intelligent Interactive Systems
@zbucinca - Researcher on Human-AI Interaction
@Samuelmoore - Researcher on scholarly communication
=> More informations about this toot | View the thread
Our paper is now published and openly available at: https://dl.acm.org/doi/10.1145/3640457.3688194
=> More informations about this toot | View the thread
22% acceptance rate at #recsys24 and 913 in-person participants
=> More informations about this toot | View the thread
Just added a few German AI research centers—BIFOLD, Tübingen AI Center and SCaDS.AI—to @ResearchOrgs! The process is super simple: just fill out a form and in 2-3 weeks your submission gets approved and goes live.
=> More informations about this toot | View the thread
Our short paper "Can editorial decisions impair journal recommendations? Analysing the impact of journal characteristics on recommendation systems" has been accepted at #RecSys2024 🎉
We analyzed how editorial decisions can impact commonly used journal recommendation algorithms. The results can also help to prevent bias in these algorithms towards e.g. mega journals.
=> More informations about this toot | View the thread
=> This profile with reblog | Go to elias@fediscience.org account This content has been proxied by September (3851b).Proxy Information
text/gemini