Sitzmann thread (3/5): mutual information supervision for interpretable features
vincesitzmann · x · 2026-10-20
Parts 2-3 of Vincent Sitzmann's thread: representation learning has lacked theory since contrastive learning, with objectives like DINO and JEPA acting as black magic. His team measures mutual information between two mask instantiations of an image with a conditional generative model, then uses that estimate to directly supervise the dot product of the masked regions' features, yielding interpretable representations.
More from Research
- Turing Post's Must-Read Papers of the Week: Cortex, Librarian Agents, Mosaic and 12 More — TheTuringPost · 2026-10-20
- MTSQL-R1 Targets the Short-Horizon Habits That Break Multi-Turn Text-to-SQL — burkov · 2026-10-20
- Robotics take: IK guarantees no precision edge, hardware—not the solver—is the bottleneck — ChongZzZhang · 2026-10-20
- MPC vs RL for robot tracking: the precision bottleneck is hardware, not the solver — ChongZzZhang · 2026-10-20
- Developer Spends 1.6M Tokens Having Codex Benchmark MPC vs RL for Arm Tracking — ChongZzZhang · 2026-10-20
- Terminal Bench 4 audit: 44 of 65 tasks cheatable, Sonnet 5.5 hits 89.2% after fixes — scaling01 · 2026-10-20