Three findings from Isola's new paper: global map, text-to-image translation, orthogonality
phillip_isola · x · 2026-10-10
Part 3 of Phillip Isola's thread, summarizing the three findings of his team's new paper against the previously listed PRH criticisms:
- A global map aligns image and text embeddings well (not just local neighborhoods);
- The alignment is strong enough to enable reasonable text-to-image translation;
- The map is orthogonal (with centering and normalization), meaning the two spaces share geometry up to rotation and scale.
These are the core conclusions of the "Shared Geometry as a Rosetta Stone" paper; details are in part 1 of the thread.
More from Research
- Mathematician John Urschel solved a decades-old problem with GPT Astra, staying optimistic on AI research — soumitrashukla9 · 2026-10-10
- State space models bring recurrence back with linear cost and transformer-era tricks — anshulkundaje · 2026-10-10
- IROS 2026 Best Paper: humanoid robot learns tennis rallies from 5 hours of human motion data — qinzytech · 2026-10-10
- AC2 lets you train custom decision models, lifting toxicity-detection F1 from 0.482 to 0.677 — rhythmrg · 2026-10-10
- KordingLab renames Plan Your Science, keeps free AI science-planning tool for researchers — KordingLab · 2026-10-10
- QuestionFirst's AI mentor critiques research plans, not with praise but 'which result would change your mind?' — KordingLab · 2026-10-10