Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 6 of 6 for “"Masked Diffusion"”.
-
Understanding theoretical and empirical foundations of diffusion models
… that have emerged over the past decade, diffusion models emerged as a powerful and principled framework, achieving state-of-the-art performance for continuous domains such as images and offering advantages like non-sequential reasoning and inference efficiency in discrete domains like …
-
Block-wise KV caching for uniform diffusion language models
Diffusion language models offer a parallel alternative to autoregressive generation: rather than predicting tokens left-to-right, they iteratively refine an entire sequence from noise in a fixed number of steps. (Austin et al., 2021; Hoogeboom et al., 2021). Among these, uniform diffusion models …
-
Probability and Representation in Neural Systems: Dynamics, Learning, and Computation
… train token prediction and unmasking order in a masked diffusion language model with on-policy reinforcement learning, improving parallel-generation accuracy--efficiency tradeoffs. Part II of this dissertation examines how learning shapes the internal representations supporting those …
-
Structure in Prediction: Mapping Large Language Model Prediction to Symbolic Structure
… the model’s objective (autoregressive vs. masked diffusion) and the sequential nature of the data. Together, these studies support a unifying principle: whether provided by the data representation, the training objective, or an external prior, structure governs the efficiency and …
-
Discrete Diffusion for Text Infilling
… parallelizability and flexible control. Discrete diffusion models offer a promising alternative paradigm, leveraging iterative refinement and potentially enabling bidirectional context use, parallel generation, and flexible prompting. However, existing discrete text diffusion models typically …
-
Gumbel distillation for parallel text generation
The slow, sequential nature of autoregressive (AR) language models has driven the adoption of parallel decoding methods. However, these non-AR models often sacrifice generation quality as they struggle to model the complex joint distribution of token sequences. To narrow this performance gap, this …