Scaling Online Optimal Transport Minibatches for Flow Matching
A scalable online entropic OT method stabilizes flow-matching training with larger source-target couplings.
Stories per day · 30 days
3 stories, newest first · one card per story
Stories tagged #training-stability
StableVQ: Practical Guidelines for Stable Vector-Quantized Tokenizer Training
StableVQ provides practical guidelines for stable Vector-Quantized tokenizer training, addressing instability between encoder and codebook optimization.
Learning Length-Extrapolatable Recurrent Models
Researchers propose Credit Stabilization through Time, a training method letting recurrent models extrapolate up to 128x their training length.