ConceptNAR模型 / Non-Autoregressive Models
Non-Autoregressive Generation
A class of text generation methods that allow simultaneous or iterative modification of entire sequences, differing from autoregressive models that generate tokens sequentially. These methods enable parallel decoding or iterative refinement, though inference typically requires repeated computation over the full sequence.
Timeline (last 90 days)
Oct 23
Non-autoregressive generative models assume that positions or levels are conditionally independent to some degree, so that multiple outputs can be produced at once or hierarchically in parallel, but they often lose fluency because dependencies between tokens are ignored.
Unverified50%