3 papers
cs.LG2026
DreamingGoose: Staged Distillation from Autoregressive Transformers to Bidirectional Recurrent Diffusion Language Models
Julian Boesch, Andrew Wee, Alexander Stranzl
Pretrained autoregressive Transformers represent a large sunk investment in compute. Existing conversion methods reuse that investment by changing either the architecture (attentio…
cs.LG2026
Anatomy of Associative Recall in Fixed-State Recurrences: A Matched-State Decomposition, an Interference Wall, and a Curriculum That Breaks It
Julian Boesch, Andrew Wee
Fixed-state recurrences--linear attention and state-space models--are reported to lag behind attention on associative recall, but whole-architecture comparisons cannot say which in…
cs.DC2024
DistriFS: A Platform and User Agnostic Approach to File Distribution
Julian Boesch
In an age where the distribution of information is crucial, current file sharing solutions suffer significant deficiencies. Popular systems such as Google Drive, torrenting and IPF…