Embedding probability measures into a Hilbert space before learning a sparse dictionary turns training from repeated Wasserstein distance solves into one autoencoder optimization, cutting reconstruction loss by roughly two to three orders of magnitude against Wasserstein dictionary learning under the same time budget.
Treating an existing archive of interventions and their already-observed outcomes as a fixed offline dataset, the work uses a vision-language model to judge whether an archived outcome matches a natural-language description and uses that judgment as the sole training reward to learn a language-to-intervention mapping for a xenobot, a synthetic multicellular construct with no nervous system; the mapping generalizes to entirely new instructions, reaching 80.0% held-out accuracy on archive data withheld from training versus a 66.7% chance baseline and matching a network trained directly on ground-truth labels.
You’re all caught up