Compute Has Shape: Self-Morphing the Internal Computation of Frozen LLMs
arXiv:2610.38382v1 Announce Type: new Abstract: Test-time scaling improves large language models through additional computation, but how to adapt the topology of internal computation to individual inputs remains underexplored. Existing approaches largely fix recurrent or parallel architectures during training, or reconfigure frozen models along a single structural axis. We introduce SelfMorph, which makes computational shape an input-dependent inference-time decision. It unifies the original fo…