[2605.23872] Training-Free Looped Transformers
Abstract:We introduce training-free looped transformers, in which a lightweight inference-time wrapper loops a contiguous mid-stack block of layers of a frozen checkpoint without additional fine-tuning, continued training, or architectural changes. Unlike prior looped transformer methods that train with the looped structure end-to-end, we retrofit recurrence onto pretrained models at test time. We show that naive block reapplication usually degrades performance, highlighting the importance of the loop application strategy. Motivated by viewing a pre-norm transformer block as a forward Euler step on an ODE, we instead treat looping as a refinement of the same approximation, replacing one large update with smaller damped sub-steps. Across seven dense, sparse MoE, and MLA+MoE model families, our method improves Qwen3-4B-Instruct by +2.64 pp on MMLU-Pro, Qwen3-30B-A3B-Instruct by +1.14 pp on CommonsenseQA, and Moonlight-16B-A3B-Instruct by +1.20 pp on OpenBookQA.
View PDF HTML (experimental) Abstract:We introduce training-free looped transformers, in which a lightweight inference-time wrapper loops a contiguous mid-stack block of layers of a frozen checkpoint without additional fine-tuning, continued training, or architectural changes. Unlike prior looped transformer methods that train with the looped structure end-to-end, we retrofit recurrence onto pretrained models at test time. We show that naive block reapplication usually degrades performance, highlighting the importance of the loop application strategy. Motivated by viewing a pre-norm…
saved by
related reading
- Do Language Models Need Sleep? Offline Recurrence for Improved Online Inferencearxiv.org
- [2607.16051] Loop the Loopies!arxiv.org
- A Mathematical Framework for Transformer Circuitstransformer-circuits.pub
- LoRA Without Regret - Thinking Machines Labthinkingmachines.ai
- GitHub - huskydoge/Awesome-Loop-Models: A curated list of papers and selected technical blogs on Loop Models.github.com
- [2607.13491] DeepLoop: Depth Scaling for Looped Transformersarxiv.org
- The Annotated Transformernlp.seas.harvard.edu
- [2604.11791] A Mechanistic Analysis of Looped Reasoning Language Modelsarxiv.org
- Looped Transformers as Programmable Computersarxiv.org
- All About Transformer Inferencejax-ml.github.io
- Fast Inference from Transformers via Speculative Decodingarxiv.org
- Getting Caught Up to Modern LLM Research | Samarth Goeldev.samarthgoel.com