LoopFormer: Elastic-Depth Looped Transformers for Latent Reasoning via Shortcut Modulation
This work addresses the need for controllable and budget-aware large language models, offering an incremental improvement by making looped architectures adaptive to compute constraints.
The paper tackled the problem of enabling looped Transformers to flexibly adapt computational depth under variable compute budgets, introducing LoopFormer with a shortcut-consistency training scheme that achieved robust performance on language modeling and reasoning benchmarks while scaling gracefully with additional budget.
Looped Transformers have emerged as an efficient and powerful class of models for reasoning in the language domain. Recent studies show that these models achieve strong performance on algorithmic and reasoning tasks, suggesting that looped architectures possess an inductive bias toward latent reasoning. However, prior approaches fix the number of loop iterations during training and inference, leaving open the question of whether these models can flexibly adapt their computational depth under variable compute budgets. We introduce LoopFormer, a looped Transformer trained on variable-length trajectories to enable budget-conditioned reasoning. Our core contribution is a shortcut-consistency training scheme that aligns trajectories of different lengths, ensuring that shorter loops yield informative representations while longer loops continue to refine them. LoopFormer conditions each loop on the current time and step size, enabling representations to evolve consistently across trajectories of varying length rather than drifting or stagnating. Empirically, LoopFormer demonstrates robust performance on language modeling and reasoning benchmarks even under aggressive compute constraints, while scaling gracefully with additional budget. These results show that looped Transformers are inherently suited for adaptive language modeling, opening a path toward controllable and budget-aware large language models.