Joint‑Embedding vs Reconstruction: Provable Benefits of Latent Space Prediction for Self‑Supervised Learning

NeurIPSSpotlight2025

Authors: Hugues Van Assel, Mark Ibrahim, Tommaso Biancalani, Aviv Regev, Randall Balestriero
Venue: NeurIPS 2025
Track: Spotlight

TL;DR

Reconstruction and joint-embedding have emerged as two leading paradigms in Self‑Supervised Learning (SSL)

Opening excerpt from the authors’ abstract. source

Read the paper

OpenReview

← All NeurIPS 2025 Spotlight papers · Browse the whole archive