Showing cs.LGShow all
2 papers · 1 filter
cs.LG2026
Placement Is Free, Composition Is Not: The Latin Square as a Provably-Balanced Construction for Heterogeneous Sequence-Mixer Stacks
Taebong Kim, Youngsik Hong, Minsik Kim +3
Since GPT, most Transformers have repeated the same attention mechanism at every layer. Yet this design is largely a convention rather than a tested conclusion. When multiple seque…
cs.LG2026
The Mask Is Not the Model: Auditing Prefix Invariance in Attention, State-Space, and Hybrid Sequence Models
Taebong Kim, Youngsik Hong, Minsik Kim +3
Hybrid sequence models must satisfy prefix invariance: representations at position t must not depend on future inputs, yet this is rarely verified. We formalize prefix invariance a…