Adaptive guidance trajectories learned via PPO outperform fixed-scale CFG on controllability-quality balance in three controlled NLP generation tasks with discrete diffusion models.
arXiv preprint arXiv:2307.03214 , year=
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 2verdicts
UNVERDICTED 2roles
background 1polarities
background 1representative citing papers
Activation Addition steers language models by adding contrastive activation vectors from prompt pairs to control high-level properties like sentiment and toxicity at inference time without training.
citing papers explorer
-
Guidance Is Not a Hyperparameter: Learning Dynamic Control in Diffusion Language Models
Adaptive guidance trajectories learned via PPO outperform fixed-scale CFG on controllability-quality balance in three controlled NLP generation tasks with discrete diffusion models.
-
Steering Language Models With Activation Engineering
Activation Addition steers language models by adding contrastive activation vectors from prompt pairs to control high-level properties like sentiment and toxicity at inference time without training.