Label-free optimal transport couplings push CFG flows to 1.91 FID on ImageNet-256
serrjoa · x · 2026-10-18
A new preprint introduces Global Transport (GT) couplings for classifier-free guided flows: a simple label-free optimal transport coupling that beats both class-conditional OT and independent couplings once the CFG scale is pushed up, reaching 1.91 FID on ImageNet-256.
Background: minibatch-OT — matching training examples to closest noise samples in each batch — improves unconditional flow models but is impractical for conditional ones. The counterintuitive fix: ignore conditioning entirely during the matching, and performance with CFG still improves.
More from Research
- Eli Ben-Sasson: wait 3-5 years before trusting lattice crypto against AI attacks — examachine · 2026-10-18
- MIT lets a Claude-powered agent freely "play" on a simulated island for 30 hours — and finds play expands its capabilities — LuizaJarovsky · 2026-10-18
- New paper measures model capability via how models solve tasks, not success rates — maksym_andr · 2026-10-18
- Economy of Minds: local incentives may replace central managers in multi-agent AI — pliang279 · 2026-10-18
- Ropedia Super Week: world model W0, dexterous bimanual controller DONIC and more — liuziwei7 · 2026-10-18
- Google Research builds AI-agent-controlled wet lab with autonomous molecular design loop — Puzzleheaded-King584 · 2026-10-18