← All works
Adaptive depth

T-LoopFormer

T-LoopFormer: Token-Level Elastic-Depth Looped Transformers with Dynamic Routing

Mingqian Yu, Wenpeng Zhang, Shaobo Cui, Peilin Zhao

Token-level elastic depth for looped transformers: a token-choice router lets each token pick its number of loop iterations from its hidden state, and a recursion-wise KV cache keeps a separate cache per loop for active tokens.

Inside the method

Token → router → [ Shared block ] × rᵢ (per-token depth)

Simplified conceptual schematic. Consult the paper for the complete architecture.

Recurrence family
Adaptive depth
Depth control
Token-choice routing
KV / state strategy
Recursion-wise KV cache per loop

Reading note

Uses LoopFormer as a baseline. Reports robust language modeling and zero-shot reasoning with the lowest decoding latency among compared models.

Sources checked 2026-10-28. This catalog does not imply independent reproduction.

Related works

Cite this work

@misc{tloopformer2026,
  title = {T-LoopFormer: Token-Level Elastic-Depth Looped Transformers with Dynamic Routing},
  author = {Mingqian Yu and Wenpeng Zhang and Shaobo Cui and Peilin Zhao},
  year = {2026},
  eprint = {2609.15160},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.15160}
}

Suggest a correction on GitHub ↗