At the end of 2025 will a DINO-based algorithm still be SOTA for self-supervised learning in vision?
3
90Ṁ40
Jan 2
43%
chance

DINO is an algorithm for self-supervised learning on images. Since its publication it's been one of the most popular algorithm for SSL, and DINOv2 is SOTA or near-SOTA on several different benchmarks. At the end of 2025 will an algorithm based on DINO still be the best? For market resolution I will use SSL classification on ImageNet.

AI
Technical AI Timelines
Get
Ṁ1,000
to start trading!
Sort by:

I think the "vision transformers need registers" paper really improves the quality for Dino v2 on things like object discovery

People are also trading

SOTA AI at EOY 2026 a reasoning model?
+10% 1d81% chance
Will Transformer based architectures still be SOTA for language modelling by 2026?
91% chance
Will there be an open source, uncensored AI image generator with the same or greater quality as DALLE-3 by end of 2025?
94% chance
What will be true of OpenAI's Sora* model, at the end of 2025? [*see description]
Will a transformer based model be SOTA for video generation by the end of 2025?
82% chance
By the end of 2025, will AI models, as Dario Amodei speculates, be able to “replicate and survive in the wild”?
9% chance
What will be the PyTorch of AI engineering at the end of 2025?
Will open-source AI win (through 2025)?
31% chance
Will the state-of-the-art AI model use latent space to reason by 2026?
13% chance
Which AI will be the best at the end of 2025?
© Manifold Markets, Inc.•Terms•Privacy