MCMC search trajectories, pruned to improving transitions, are distilled into an autoregressive model that extrapolates sequence scores beyond both the training data and the search itself.
Make this {positive, negative }:
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Learning Extrapolative Sequence Transformations from Markov Chains
MCMC search trajectories, pruned to improving transitions, are distilled into an autoregressive model that extrapolates sequence scores beyond both the training data and the search itself.